diff --git a/.gitattributes b/.gitattributes index 72fb9d0eff182a8954301913e727b06149e056c3..af3c6f3538183304754fd7e85ccc8bbc88b95a04 100644 --- a/.gitattributes +++ b/.gitattributes @@ -201,3 +201,9 @@ ptFPT4oBgHgl3EQf7zXe/content/2301.13206v1.pdf filter=lfs diff=lfs merge=lfs -tex ytE1T4oBgHgl3EQfQwPe/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text WNAyT4oBgHgl3EQfu_m7/content/2301.00624v1.pdf filter=lfs diff=lfs merge=lfs -text I9FAT4oBgHgl3EQfux7Z/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text +w9FJT4oBgHgl3EQfgiwX/content/2301.11561v1.pdf filter=lfs diff=lfs merge=lfs -text +zNE4T4oBgHgl3EQfAQs4/content/2301.04841v1.pdf filter=lfs diff=lfs merge=lfs -text +fNE_T4oBgHgl3EQf2BxV/content/2301.08338v1.pdf filter=lfs diff=lfs merge=lfs -text +x9E3T4oBgHgl3EQflwp-/content/2301.04610v1.pdf filter=lfs diff=lfs merge=lfs -text +h9E0T4oBgHgl3EQfpgF0/content/2301.02540v1.pdf filter=lfs diff=lfs merge=lfs -text +c9E3T4oBgHgl3EQfeQoX/vector_store/index.faiss filter=lfs diff=lfs merge=lfs -text diff --git a/29E3T4oBgHgl3EQfoAqh/content/tmp_files/2301.04630v1.pdf.txt b/29E3T4oBgHgl3EQfoAqh/content/tmp_files/2301.04630v1.pdf.txt new file mode 100644 index 0000000000000000000000000000000000000000..64bfa918973923ab268a6018b94e09587f7bf973 --- /dev/null +++ b/29E3T4oBgHgl3EQfoAqh/content/tmp_files/2301.04630v1.pdf.txt @@ -0,0 +1,1590 @@ +ShadowNav: Crater-Based Localization for Nighttime +and Permanently Shadowed Region Lunar Navigation +Abhishek Cauligi* +abhishek.s.cauligi@jpl.nasa.gov +R. Michael Swan* +robert.m.swan@jpl.nasa.gov +Hiro Ono +masahiro.ono@jpl.nasa.gov +Shreyansh Daftry +shreyansh.daftry@jpl.nasa.gov +John Elliott +john.o.elliott@jpl.nasa.gov +Larry Matthies +lhm@jpl.nasa.gov +Deegan Atha +deegan.j.atha@jpl.nasa.gov +Jet Propulsion Laboratory, California Institute of Technology +Pasadena, CA 91109, USA +Abstract—There has been an increase in interest in missions +that drive significantly longer distances per day than what +has currently been performed. +For example, Endurance-A +proposes driving several kilometers a day in order to reach +its target traverse of 2000 km in 4 years. Additionally, some +of these proposed missions, including Endurance-A and rovers +for Permanently Shadowed Regions (PSRs) of the moon, re- +quire autonomous driving and absolute localization in darkness. +Endurance-A proposes to drive 1200 km of its total traverse at +night. The lack of natural light available during these missions +limits what can be used as visual landmarks and the range +at which landmarks can be observed. In order for planetary +rovers to traverse long-ranges, onboard absolute localization is +critical to the rover’s ability to maintain its planned trajectory +and avoid known hazardous regions. Currently, the localization +performed onboard rovers is relative to the rover’s frame of +reference and is performed through the integration of wheel +and visual odometry and inertial measurements. To accomplish +absolute localization, a “ground-in-the-loop” (GITL) operation +is performed wherein a human operator matches local maps +or images from onboard with orbital images and maps. This +GITL operation places a limit on the distance that can be +driven in a day to a few hundred meters, which is the distance +that the rover can maintain acceptable localization error via +relative methods. Previous work has shown that using craters +as landmarks is a promising approach for performing absolute +localization on the moon during the day. +In this work we +present a method of absolute localization that utilizes craters +as landmarks and matches detected crater edges on the surface +with known craters in orbital maps. We focus on a localization +method based on a perception system which has an external +illuminator and a stereo camera. While other methods based +on lidar exist, lidar is not currently planned for deployment +on the current proposed nighttime and PSR missions. In this +paper, we evaluate (1) both monocular and stereo based surface +crater edge detection techniques, (2) methods of scoring the +crater edge matches for optimal localization, and (3) localization +performance on simulated Lunar surface imagery at night. We +demonstrate that this technique shows promise for maintaining +absolute localization error of less than 10 m required for most +planetary rover missions. +TABLE OF CONTENTS +1. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +1 +2. RELATED WORKS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +2 +*Abhishek Cauligi and R. Michael Swan contributed equally to this work. +978-1-6654-9032-0/23/$31.00 ©2023. California Institute of Technology. +Government sponsorship acknowledged. +The research was carried out at the Jet Propulsion Laboratory, California +Institute of Technology, under a contract with the National Aeronautics and +Space Administration (80NM0018D0004). +3. APPROACH. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +3 +4. DATASETS OVERVIEW . . . . . . . . . . . . . . . . . . . . . . . . . . . . +5 +5. CRATER DETECTION PERFORMANCE . . . . . . . . . . . +6 +6. LOCALIZATION PERFORMANCE . . . . . . . . . . . . . . . . . +8 +7. CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +10 +ACKNOWLEDGMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +10 +REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +11 +BIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +12 +L +R +Figure 1: The ShadowNav localization algorithm per- +forms absolute localization for a Lunar rover mission +located at the red position in the left image by matching +known craters from (left) an orbital map against (right) +detected craters from the rover stereo cameras. +1. INTRODUCTION +Long-range Lunar navigation, and specifically navigating +within darkness, has gained a significant amount of traction +recently. For example, missions to Permanently Shadowed +Regions (PSRs) of the moon have been proposed such as +the VIPER mission [1], [2] and the Lunar Polar Volatiles +Explorer mission concepts. Furthermore, there are missions +that have proposed driving during the Lunar night in order +to traverse longer distances. For example, the new Decadal +Survey [3] recommends the Endurance-A Lunar rover mis- +sion should be implemented as a strategic medium-class +mission as the highest priority of the Lunar Discovery and +Exploration Program. The Endurance-A rover proposal plans +to drive 2000 km in the South Pole-Aitken (SPA) Basin to +collect +100 kg of samples, which would be delivered to +Artemis astronauts. This mission concept study [4] identified +several key capabilities required to complete this mission +which are: (1) Endurance will need to drive 70% of its total +distance during the night to enable daytime hours dedicated +to science and sampling. (2) The mission will require on- +board autonomy for the majority of its operations, while the +1 +arXiv:2301.04630v1 [cs.RO] 11 Jan 2023 + +Orbital Image Generation +Stereo Images Generation +Crater Edge +Detection +Local-to-Global +Transform +Particle Filter +Step +Compute Q-Score +Orbital Image Generation +Particle Filter +Disparity +Generation +Disparity Hole +Filler +Figure 2: Schematic of the ShadowNav algorithm proposed to perform absolute localization on the Moon. A particle +filter is used to match craters detected by the rover stero cameras with known craters from an orbital map. +ground only handles contingencies. (3) Global localization +is necessary to maintain an error of <10 m relative to orbital +maps. +At present, existing rovers perform onboard localization rel- +ative to their own reference frame. +This is accomplished +by using wheel and visual odometry and inertial measure- +ments. Absolute localization is performed periodically with +a “ground-in-the-loop” (GITL) operation. This is acceptable +for current driving distances which are a few hundred meters +a day. Existing relative localization has around 2% drift and +therefore can only drive at most 500 m before the error will +be larger than 10 m. In order to traverse longer distances, on +the order of several kilometers a day proposed by missions +such as Endurance-A, autonomous absolute localization be- +comes critical. At present the Lunar surface does not have +continuous communication with Earth. Therefore, having to +perform several GITL operations for absolute localization in +a day will significantly reduce the distance that can be driven. +The lack of frequent absolute localization for the rover would +lead to errors greater than the maximum 10 m localization +error which would present significant risks to the mission +through deviations from the desired trajectory and risk for +unidentified obstacles. +Craters as landmarks have been shown to be promising for +absolute localization on the Moon [5], [6]. However, the lack +of natural light available while driving within a PSR or during +the Lunar night limits what can be used as a landmark and the +range at which the landmarks can be observed. Using craters +is still promising as the average distance between craters of +≥10 m in diameter is 100 m on terrain with relatively fresh +craters and +10 m on terrain with old craters [7]. +Addi- +tionally the Lunar Reconnaissance Orbiter Camera (LROC) +provides digital elevation models (DEMs) with a resolution +between 0.5 m-5 m per pixel [8] and there are some DEMs +within PSRs [9]. +In this work, we propose using a stereo camera with an +illuminator positioned below the stereo camera in order to +detect crater rims within the darkness. The use of such an +illuminator is motivated by the Endurance-A mission concept +study [4], which proposes the use of a stereo camera with +an illumination source as the perception system for a rover +operating in darkness. +Global localization is then accom- +plished by matching the detected crater rims against known +craters from an orbital image as shown in Figure 2. To handle +the uncertainty and nonlinearity of the crater rim detection +model, we utilize a particle filter with a novel Q-Score metric +for ranking potential crater matches in order to estimate the +absolute position of the rover within an orbital map. This +paper demonstrates the initial results of both crater detection +within darkness and absolute localization within simulation +which are the results of the first two years of a planned +three year effort to validate this approach. Work is ongoing +to collect and validate this approach in a real-world Lunar +analogue test location. +Statement of Contributions: This paper presents an approach +to absolute localization on the Moon that can be performed +while a rover is in darkness, such as within a PSR or during +the Lunar night. +The main contributions of the work as +summarized below: +1. We developed a simulator based on Blender [10] which +renders simulated surface stereo imagery of the Lunar sur- +face in darkness located within a known orbital position. +The rendering process utilizes the Hapke lighting model for +more accurate surface reflectance as well as DEMs captured +by LROC for realistic crater distributions. +2. We evaluated different crater-edge detection techniques +and demonstrate a method which captures 80% of the leading +crater arc at 10 m and can detect crater arcs out to 20 m. +3. We present a method to localize a rover within an orbital +map using surface crater-edge detections and known orbital +craters based on a particle filter and a metric we call the Q- +Score which is detailed in Section 3. +4. We demonstrate our absolute localization technique can +achieve less than 2 m absolute error with an assumed odome- +try drift of 2% and an initial 3-sigma uncertainty of 3 m. +2. RELATED WORKS +Absolute localization on planetary surfaces is critical for +expanding the range rovers can travel in a day and over the +course of a mission and there have been many previous works +that investigate this problem. There have been techniques +proposed for the Martian surface. Works such as [11], [12] +consider far range and horizon features which are at ranges +that are beyond what is expected can be seen in the dark. +2 + +Figure 3: Figure demonstrating the impact of placement +of light source on crater rim shadows. +Left: Sample +render of a crater with light source even with camera. +Right Sample render of a crater with light source below +the camera. +[13] proposes a technique on the Martian surface for absolute +localization that uses rocks and DEMs surface features. +In our work, we focus on the problem of global localization in +darkness which is relevant for permanently shadowed regions +of the moon, for which there has been a surge of interest in +conducting scientific measurements and activities [14]. Our +solution approach is inspired by a host of recent works that +seek to leverage orbital maps for global rover localization in +these shadowed regions. In [13], the authors propose a local- +ization procedure that matches an observed rover image with +an orbital map, but this approach neglects the rover motion +model and yields a deterministic estimate of the robot belief. +A purely data-driven approach is presented in [15], wherein +a convolutional neural network is trained on synthetic data to +match the rover observations with orbital imagery. Closest +to our approach, [16] presents a particle filtering technique +to compare rover monocular camera imagery with orbital +imagery and uses a Siamese neural network approach to +assign each particle a likelihood weight. The authors in [6] +propose a similar approach for Lunar absolute localization +known as LunarNav. However, LunarNav focuses on the day- +time localization problem and therefore considers different +methods of crater matching that rely on greater knowledge of +the surface geometry than available in the nighttime case. +3. APPROACH +In this work, we propose an absolute localization approach +which utilizes a crater’s leading edge as landmarks for local- +ization. The end result of this approach will be an estimated +position and uncertainty within the orbital frame. At present, +this approach only considers position localization. +Rover +orientation is assumed to be given by a star tracker which +can compute orientation in three dimensions from celestial +measurements. Our approach consists of two primary com- +ponents: +1. A leading-edge crater detection methodology for use with +a Lunar rover equipped with a stereo camera system and +illumination source. +2. A particle filter for computing a position belief based on +a score computed based on the association of crater edges +and known orbital ground truth craters, which we call the Q- +Score, and the robot motion model. +A. Surface Crater Detection +In order to identify craters on the surface, the system was de- +signed to be used in conjunction with a perception system that +contained a stereo camera and an illumination source. This +perception system was configured where the illumination +source was beneath the stereo camera. Examples of simulated +images with the light at the same height as the cameras and +the light positioned beneath the cameras are in Figure 3. It +was observed that placing the illumination source below the +camera results in a shadow at the leading edge of a negative +obstacle. Furthermore, offsetting the light with the cameras +reduced the impact of the Hapke model washing out some of +the surface texture. Further details on the Hapke model and +its impact on surface terrain are provided in Section 4. +Here, we first review the three different techniques studied in +this work for detecting a crater’s leading edge: (1) a method +of detecting jumps within stero disparities, (2) a Canny edge +detector used to find the shadow on the leading edge, and (3) +a convolutional neural network (CNN)-based edge detector +that uses both the monocular and disparity image as input. +1. Stereo Disparity Discontinuity Method The first approach +for leading edge crater detection relies on detecting discon- +tinuities within the stereo disparity image. To accomplish +this, the stereo disparity image must first be generated using +methods such as the JPLV algorithm [17] or the Semi-Global +Block Matching (SGBM) approach [18], among others. To +account for the low contrast that may be present in the Lunar +rover case, Contrast Limited Adaptive Histogram Equaliza- +tion (CLAHE) is first run on the input images prior to running +stereo. CLAHE is an adaptive histogram equalization and +operates on sub-regions of an image which allows more +consistent equalization across different lighting conditions +within an image. This is useful for this application as there is +a light-to-dark gradient from near-to-far within the images. +The resulting disparity image is then scanned column-by- +column and, when the difference between any two disparities +is greater than some pre-defined threshold, the larger column +index is marked as a crater edge. +Further, any numerical +issues stemming from stereo holes are accounted for by +omitting any pixels with spurious values during comparison. +2. +Canny Edge Detector Method For sensor configura- +tions that contain an illuminator located beneath the stereo +cameras, shadows appear on the leading edge of negative +obstacles. In such cases, a Canny edge detector can be used +to distinguish the stark contrasting dark line along the rim. +In this work, the Canny edge detector from OpenCV [19] is +used to find these shadows. +3. +CNN-Based Edge Detector Method The Holistically- +Nested Edge Detection (HED) approach presents a CNN- +based deep learning based method for leading edge crater +detection [20]. +This method uses the HED approach and +can be performed by directly using the publicly released +neural network weights. HED is capable of performing both +monocular and stereo depth based edge detection. For HED +to perform edge detection within a depth image, it generates a +three channel image that contains horizontal disparity, height +above ground, and angle of the local surface normal with the +inferred direction of gravity. The RGB and depth predictions +of the CNN are then merged to generate the desired output. +Positive Obstacle False Positive Rejection—One shortcoming +of the aforementioned leading edge crater detection approach +is the susceptibility of false positive cases in the presence +of positive obstacles. In order to account for this positive +3 + +Algorithm 1 Q-Score Computation +Require: Belief bt +i, set of crater observations {zt +0,rover, ..., zt +m,rover}, +set of ground truth craters {ct +0,world, ..., ct +ℓ,world}, positive value +ε +1: +Qinc ← ε +2: +for i = 1, . . . , m do +3: +zt +0,world ← rover to world(zt +0,rover) +4: +dcr ← min ∥cj,world − zt +0,world∥ +5: +Qinc ← Qinc + dcr +6: +end for +7: +Qscore ← min +� +1, ( 1 +mQinc)−1� +8: +return Qscore +Algorithm 2 ShadowNav Particle Filtering Algorithm +Require: Initial belief distribution (µ0, Σ0), number of particles +Ns, number of effective particles threshold Neff,thresh +1: +{b0 +1, ..., b0 +Ns} ← sample beliefs(µ0, Σ0) +2: +{w0 +1, ..., w0 +Ns} ← {1, ..., 1} +3: +t ← 1 +4: +while particle filter running do +5: +{zt +0, ..., zt +m} ← get observations() +6: +{qt +1, ..., qt +Ns} ← {0, ..., 0} +7: +for i = 1, ..., Ns do +8: +bt +i ← propagate sample(bt−1 +i +) +9: +qt +i ← log Q score(bt +i, {zt +0, ..., zt +m}) +10: +end for +11: +qt +min ← min(qt +1, ..., qt +Ns) +12: +for i = 1, ..., Ns do +13: +wt +i ← wt−1 +i ++ qt +i − qt +min +14: +end for +15: +Neff ← compute Neff(wt +1, ..., wt +Ns) +16: +if Neff ≤ Neff,thresh then +17: +{bt +1, ..., bt +Ns} ← resample beliefs({bt +i}Ns +i=1, {wt +i}Ns +i=1) +18: +{wt +1, ..., wt +Ns} ← {1, ..., 1} +19: +end if +20: +t ← t + 1 +21: end while +obstacle issue, the detected edge points are passed through +a filter that removes points which have hits on the far side of +the crater edge with a detected negative or flat slope. Detected +edge points are kept only if, within the region directly beyond +the detected edge, there exists a positive slope or if there is +not enough stereo to accurately compute the slope. Thus, the +case of a detected positive slope is assumed to correspond +to the rising edge of the crater under the assumption that +the detected edge is the leading edge of a negative obstacle. +Alternatively, a detected edge is also retained if the far edge +is not captured due to low light conditions, as this is assumed +to be an indication of the presence of a large crater. +B. Particle Filter +Here, we provide an overview of the proposed ShadowNav +particle filtering approach. First, we provide further details +on the Q-Score metric that is used in the belief update step. +Q-Score—The Q-Score provided the measurement probabil- +ity of some position belief based on rover frame observations +and an orbital map. The procedure for computing the Q-Score +is given in Algorithm 1. The algorithm takes as input a given +belief bt +i, a set of m observed edges in rover frame, and a set +of ℓ ground truth crater observations to associate these mea- +surements with (Line 1). A value Qinc is initialized to some +negligibly small, positive value ε to later avoid divide-by-zero +Algorithm 3 Systematic Resampling +Require: Particles +{bt +1, ..., bt +Ns} +and +associated +weights +{wt +1, ..., wt +Ns} +1: +nt = log +� �Ns +i=1 exp(bt +i) +� +2: +{ ˜wt +0, ..., ˜wt +Ns} ← {0, ..., 0} +3: +for i = 1, ..., Ns do +4: +˜wt +i ← exp(wt +i − nt) +5: +end for +6: +{q0, ..., qNs} ← cum sum({ ˜wt +0, ..., ˜wt +Ns}) +7: +n ← 0 +8: +m ← 0 +9: +u0 ∼ U(0, +1 +Ns ) +10: while n ≤ Ns do +11: +u = u0 + +n +Ns +12: +while qm ≤ u do +13: +m ← m + 1 +14: +end while +15: +n ← n + 1 +16: +bt +n ← bt +m +17: end while +18: return {bt +0, ..., bt +Ns} +issues (Line 1). Next, for each measurement zt +i in the rover +frame, the detected edge is converted to world frame (Line 3) +and the minimum distance to an edge from the ground truth +map computed (Line 4). +The Qinc is incremented by the +distance between the observed edge and its associated ground +truth observation (Line 5). The Q-Score is computed as the +reciprocal of Qinc and a min operation is applied to ensure +that the score provided by any particular run is between 0 and +1 (Line 7). This implies that observations and belief pairs +which are less than 1 m away from ground truth will receive +the same score as those exactly 1m away from ground truth, +which is seen as acceptable given the orbital DEM resolution +and mission concept localization requirements. +In addition to the shortest distance formulation from Line 4, +additional approaches were also explored for determining +the Q-Score. One alternate approach investigated included +fitting a Gaussian normal distribution on the orbital map +crater edges and the Q-Score value was them computed based +on the intensity (i.e., distance to the computed mean) of the +point hit by observations or 0 in cases when no point was +hit. In practice, it was determined that the shortest distance +formulation provided the most robust results for use with the +particle filter and also did not require additional projection +calculations to project each belief from the orbital frame to +rover frame. +Overview—A description of the ShadowNav particle filtering +algorithm is given in Alg. 2. The algorithm takes as input a +Gaussian belief distribution (µ0, Σ0) assumed for the initial +robot position, the number of particles Ns to use in the +particle filter, and a threshold for the effective number of +beliefs Neff,thresh used to trigger resampling (Line 2). +The +filter is initialized by sampling Ns particles from the initial +belief distribution and assigning a weight of equal importance +for each particle (Lines 1-2). As common in particle filtering +implementations [21], we note that we used the log of the +weights for improved numerical stability of the weight update +step [22]. Given a new set of crater observations (Line 5), +a set of Q-Score measurements is initialized for computing +for each individual particle (Line 6). +After applying the +motion model update to each particle (Line 8), the Q-Score +for each updated particle is computed using the procedure +from Alg. 1 by comparing against the current measurements +4 + +(a) Sample of terrain +with 90◦ from cam- +era. +(b) Sample opposi- +tion effect during the +day. +(c) Sample effect of +surface reflectance at +night with an illumi- +nator. +Figure 4: The opposition effect simulated during the day +and its effect at night with an external illuminator. +(Line 9). The particle weights are then updated in log-domain +(Line 13) with a normalization step to ensure non-negative +weights (Line 11). Next, the number of effective samples Neff +at the current iteration is calculated (Line 15). A common +pitfall of particle filters is “degeneracy”, wherein the weights +{wt +i} collapse around a handful of particles and computa- +tional resources are wasted on propagating low likelihood +particles [21]. If Neff is below the threshold Neff,thresh, then +this indicates that the filter is degenerating and a resample +operation is triggered (Line 17). +Further details on the systematic resampling approach used in +this work are provided in Algorithm 3. Given a set of particles +and their associated weights, the weights are first normalized +to (0, 1] from log-domain (Lines 1-4) and the cumulative sum +of these normalized weights ˜wt +i computed (Line 6). The key +step in systematic resampling is to sample a random value +u0 from a uniform distribution inversely proportional to Ns +(Line 9) and then incrementally sample a new particle from +this “bin” of width +1 +Ns . This ensures that, after resampling, +at least one particle is retained from each +1 +Ns interval from +the previous belief distribution. +C. Surface to Orbital Crater Transformation +For every observation step, rover frame crater edges were +detected with a stereo camera pair that provided the depth, +and thus a relative position for the crater edge was saved. This +relative crater distance was added to each particle’s belief +position to form an estimate of the observed crater position +in the world frame for each particle. The orbital map was +projected to the world frame and then the shortest distance +metric noted in the Q-Score algorithm was used to determine +which particle belief positions were most likely and thus +which observed crater was the most likely one to match the +known orbital craters. +Stereo hole filling—As some crater edge detections do not +rely on depth information, not all pixels in the stero camera +depth or disparity image will have a detected depth value +and, in such cases, no relative position would be available +for matching rover observations to the orbital map. For such +observations, a simple plane fit can be carried out to fill in the +depth information. A future area of investigation includes +carrying out an improved stereo hole filling approach, in +particular using existing knowledge on what the regional +terrain looks like. +Figure 5: The trajectories used for the numerical ex- +periments are overlaid on the orbital map here with the +crater numbers in black. +A red square indicator is at +the start and a green circle indicator is at the end of +each trajectory. Trajectory 1 is in blue, trajectory 2 is +in orange, and trajectory 3 is in purple. +4. DATASETS OVERVIEW +A. Simulated Lunar Environment +At the time of writing, a Lunar dataset with images captured +in the dark with an illuminator did not exist. Therefore, to +evaluate the approach, a simulation environment was devel- +oped using the Blender software [10]. In order to simulate +images as realistically as possible, the Hapke lighting model +[23], [24], [25] was implemented. This model approximates +the Lunar surface reflectance and will simulate the “opposi- +tion effect”. This effect leads to a focused point of extreme +saturation at a location within an image where the camera ray +and light source are at zero phase angle. The Hapke lighting +model was implemented using the “old highland” parameters +of the moon provided in [26], as these most closely match the +poles of the moon where PSRs can be found. The coherent +backscattering opposition effect (CBOE) was left out of our +implementation and only the shadow hiding opposition effect +(SHOE) was implemented as it dominates most or all lighting +calculations in our use case, while CBOE has a negligible +or very small effect. Initial implementation was done using +the Open Shading Language (OSL), however not all rays are +available for calculation due to optimizations made in OSL, +so workarounds were needed to implement the Hapke light- +ing model in Blender using OSL. While this was partially +successful, it was not very robust and we had numerous +issues. +Instead of using OSL, we opted to modify the +source of Blender to add the Hapke bidirectional reflectance +distribution function (BRDF) directly into the Blender Cycles +renderer code which also reduced the render time by greater +than a factor of 2 through the use of Nvidia CUDA. +In order to represent a realistic 3D model of the sur- +face geometry, DEMs produced from LROC were utilized. +While LROC has enough resolution to resolve craters of +around 10 m, its resolution is not quite good enough for +generating smooth surface imagery. In order to have smooth +surface image renders, the DEMs from LROC were scaled +down to be 0.25 m resolution. Crater measurements in fu- +5 + +10Table 1: Table of crater sizes in crater detection dataset. +Crater +Diameter (m) +Depth (m) +1 +9.2 +1.0 +2 +9.1 +0.75 +3 +11.3 +0.84 +4 +4.4 +0.55 +5 +3.7 +0.40 +6 +8.3 +0.27 +7 +11.9 +0.44 +8 +3.9 +0.48 +9 +4.1 +0.49 +10 +2.3 +0.25 +ture discussions were based on this scaled resolution. This +scaled DEM was imported into Blender and a surface texture +was added. +The surface texture comprised of two scales +of fractal Brownian motion, which is a natural noise that +was added to the DEM in order to simulate Lunar surface +texture for stereo to utilize. +Figure 4 demonstrates three +sample renders, two in the daylight and one at night with +an illumination source from our simulation. It demonstrates +what the surface looks like in daytime conditions as well as +the effect of the Hapke model during the day with the sun +behind the camera and the effect of the illumination source. +From this it was observed that the full amount of daytime +texture is not observed during the night with an illumination +source. +B. Simulated Craters for Detection Analysis +In order to evaluate the performance of different crater de- +tection techniques, a dataset with different sized craters was +built. +This dataset was built using the simulation process +within Blender and captured stereo pair renders between 5 m +and 20 m from the front crater rim in increments of 0.1 m. +This dataset contained 10 different craters with varying sizes +and depths. The sizes of the craters within this dataset are in +Table 1 and their locations corresponding to the crater ID in +our simulated environment are marked in Figure 5. +C. Simulated Trajectories for Localization Analysis +In order to evaluate the localization performance, several +trajectories were run in the simulated environment. These +trajectories were run to generate an image every 1 m and +were designed to approach craters in different ways that +might present challenges to our filtering approach. +The +1 m observation delta was used to reduce render times of +our dataset, as rendering every 0.1 m did not result in a +significant localization performance change. An overview of +the trajectories within the orbital environment are displayed +in Figure 5. +D. Real Data of Negative Obstacles at Night +In addition to the simulated data generated, a dataset was +collected in the Arroyo, which is a dry river bed near the +NASA Jet Propulsion Laboratory. This dataset contained a +few different negative obstacles that were imaged at 5, 10, +and 15 m away from the leading edge. +This dataset was +used to validated that the stereo and crater edge detection +algorithms work on real data collected at night with an +external illuminator. +Figure 6: +Plots of different metrics evaluating crater +detection performance. Left: Plot that shows image-based +crater edge detection score versus range for all craters +evaluated. Right: Plot that shows percent of the crater +front arc detected for all craters evaluated. +Figure 7: Sample stereo results using JPLV stereo on a +sample negative obstacle. +5. CRATER DETECTION PERFORMANCE +A. Metrics +In order to evaluate the performance of surface crater detec- +tion, the dataset referenced in Section 4 was utilized. Five +different combinations of algorithms were evaluated. These +were disparity discontinuity detection within SGBM stereo, +disparity discontinuity detection within JPLV stereo, HED +using SGBM stereo, HED using JPLV stereo, and a hybrid +JPLV disparity discontinuity detection and canny edge detec- +tion approach. The hybrid discontinuity detection and canny +approach was implemented so that Canny only ran on the +portion of the image that was 10 m away or further. This +was done since it was observed the discontinuity detection +worked well in the near range but stereo began to degrade +beyond 10 m. +These algorithms were evaluated with two different metrics. +The first was an image based edge scoring method which +captures an average Gaussian probability that a detected edge +is on a ground truth crater edge. It utilizes a distance error +computed in image space as represented in Equation 1 where +Errordistpx is the pixel error from ground truth to detection, +rangegt is the known ground truth range, fl is the focal length +of camera, ss is the sensor size of the camera, and Errordist +is the error in meters of the detection. +Errordist = Errordistpx ∗ rangegt +(fl ∗ ss) +(1) +6 + +5m +10mImageBasedCraterEdgeScoreversusRange +1.0 ++ +Disparity JPLV +Disparity SGBM ++ +HED JPLV +0.8 +HEDSGBM +Disparity + Canny jPLV +0.6 +Score ++ ++ ++ +0.4 +0.2 +0.0 +6 +8 +10 +12 +14 +16 +18 +20 +GT Range (m)Percent Crater Front Arc Detected vs.Range +100 ++ +Disparity JPLV +Percent of Crater Front Arc Detected (%) +HED JPLV +Disparity SGBM +80 +HED SGBM +Disparity JPLV + Canny +60 +++ +40 +20 +0 +6 +8 +10 +12 +14 +16 +18 +20 +Range(m)(a) Ground Truth at 7 m +(b) Ground Truth at 12 m +(c) Ground Truth at 17 m +(d) JPLV Disparity + Canny at 7 m +(e) JPLV Disparity + Canny at 12 m +(f) JPLV Disparity + Canny at 17 m +(g) JPLV HED at 7 m +(h) JPLV HED at 12 m +(i) JPLV HED at 17 m +Figure 8: The efficacy of the JPLV HED approach over JPLV Disparity + Canny is demonstrated in simulations of +crater rim detection overlay samples for crater 1. +The distance error was then passed into a Gaussian. +The +Gaussian probabilities for all of the detected pixels were +summed together and normalized by number of detected +points to obtain a score. This scoring method infused ground +truth range values to remove the impact of stereo holes +and stereo range uncertainty on the projection in order to +better isolate the specific performance of the crater detection +algorithms. The sigma value for the Gaussian that was used +in these experiments was 0.25m. This was chosen because +the resolution of the DEM utilized was 0.25 m. Therefore +most detections should fall within this boundary if they are +highly accurate. +The second metric used was ”percent of +front arc detected”. In this metric, there is a ground truth +circle of the orbital crater. +Depending on the pose of the +simulated cameras, the half arc of the ground truth circle that +was nearest the simulated camera was projected into image +space. +The crater detection was then matched to the half +arc and the percentage of the half arc that was successfully +identified was determined. This metric removes the Gaussian +component from the first metric; however, it does not capture +7 + +(a) Negative obstacle 5 m away +(b) Negative obstacle 10 m away +(c) Negative obstacle 10 m away +(d) Negative obstacle 5 m away +(e) Negative obstacle 15 m away +(f) Negative obstacle 10 m away +Figure 9: Qualitative edge detections using JPLV disparity discontinuity detection and Canny hybrid on negative +obstacles on a real dataset collected in a dry river bed at night. These results demonstrate the transferability of the +crater detection algorithms from simulated data to a real environment. +false positives like the first metric. +B. Detection Results on Simulated Data +The results of running the different algorithms on the simu- +lated dataset are observed in Figure 6. There were several +notable observations from the results. The first was that the +algorithms tended to perform the best around 10 m and did +not improve as craters came closer. +This was believed to +be because as the camera gets closer to the crater, more of +the crater becomes visible and the discontinuities become +smaller. However, as the crater becomes further than 10 m, +the stereo began to degrade. +Additionally, for the hybrid +stereo and Canny technique, the Canny detection started de- +tection at 10 m and led to a significant jump in performance. +In terms of algorithm comparison, JPLV disparity disconti- +nuity performed better than SGBM disparity discontinuity +which is likely due JPLV having more holes than SGBM. +These holes at the boundary helped the disparity discontinuity +detector find a better edge. However, for HED, it performed +well with either stereo technique, likely due to its representa- +tion of depth containing height values. HED was used with +its out of the box weights from its authors. It likely could be +improved with finetuning on a Lunar dataset. +In addition to quantitative results, samples of crater rim de- +tection overlays are in Figure 8. These results were on crater +1 which is a nearly 10 m in diameter crater. Both methods +were able to detect the craters well, but JPLV HED did have +more falloff at 17 m than the Canny detector. +However, +the Canny edge detector was optimized for this environment +where as HED was a generalized detector. +Overall the +generalization of HED was extremely promising as a crater +rim detection approach. +C. Detection Results on Real Data +As described previously, data was collected from a location +with negative obstacles at night. +This dataset was used +to validate the performance of stereo and crater detection +algorithms. +Figure 7 presents a sample of 5 m and 10 m +negative obstacles and the corresponding stereo results from +JPLV. From this figure is was observed that stereo is dense up +unto the leading edge of the negative obstacle. Additionally, +at 5 m, the far edge of the negative obstacle was captured +in the disparity values. At 10 m, the far edge, did contain +some disparity values but it was sparse. +While not fully +representative of the Lunar surface, this demonstrated that +current stereo techniques do have the capability to work in +low light conditions at the ranges necessary. The data was +also used to evaluate the edge detection techniques. The JPLV +disparity discontinuity and Canny edge detection hybrid was +found to be the best on simulation data and therefore it +was used on the real data. +Figure 9 demonstrates sample +detections at different ranges. +These detection results did +contain false positives on some of the vegetation as the false +positive rejection was not run. +Vegetation is not present +on the moon, however, objects such as rocks could present +similar issues. Overall, the negative obstacle edge detection +qualitatively performs well. +6. LOCALIZATION PERFORMANCE +In this section, we provide Monte Carlo results on the per- +formance of the proposed ShadowNav filtering algorithm. +For each simulation, we analyzed the performance of the +ShadowNav filter on the basis of the following metrics: +8 + +(a) Ground truth error for traj. 1. (b) Filter covariance for traj. 1. (c) Ground truth error for traj. 2. (d) Filter covariance for traj. 2. +Figure 10: A comparison of the four proposed resampling schemes demonstrated that systematic resampling empirically +outperforms the other scheme in terms of relatively lower ground truth error and reduced uncertainty in the filter. +(a) Ground truth error. +(b) Filter covariance. +Figure 11: Monte Carlo simulations for trajectories 1–3 +demonstrated the efficacy of the Q-Score based particle +filtering approach at accomplishing global rover localiza- +tion. +(a) Traj. 1 traverse – case A. +(b) Traj. 1 traverse – case B. +Figure 12: Two Monte Carlo trials for trajectory 1 are +illustrated with the ground truth in red and the weighted +average belief µt in blue. The comparatively better per- +formance of the filter in case A (left) was due to false +positive crater rim measurements in case B (right) that led +to worse localization. +Ground truth error: We computed the weighted average +mean µt = �Ns +i=1 wt +ibt +i at time t for the filter using the particle +weights and beliefs and compute the ℓ2-distance to the ground +truth gtt, i.e., ∥µt − gtt∥2. +Particle filter uncertainty: To capture the uncertainty asso- +ciated with the current belief, we additionally computed the +weighted covariance matrix Σt = �Ns +i=1 ˜wt +i(bt +i−µt)(bt +i−µt), +where ˜wt +i are the normalized weights detailed in Alg. 3. The +metric we report at each time step was the square root of +the largest eigenvalue +� +λmax(Σt), which corresponded to the +worst case variance of the estimation error [27], [28]. +Mahalanobis distance: The final metric we computed was the +Mahalanobis distance, which measures the distance between +and the particle filter distribution and ground truth posi- +tion. We approximately computed this by fitting a Gaussian +distribution N(µt, Σt) to the particle filter distribution, for +which the Mahalnobis distance is simply a weighted ℓ2-norm +� +(µt − gtt)T (Σt)−1(µt − gtt). +A. Resampling Scheme Comparison +In this section, we compared the baseline systematic resam- +pling approach detailed in Alg. 3 against three other resam- +pling methods utilized: multinomial, residual, and stratified +(we refer the reader to [21], [29], [30] for a thorough review +of these approaches.) Figure 10 presents the ground truth +error and filter uncertainty for the four different resampling +approaches. We saw that, for the two trajectories compared +in Figure 10, systematic resampling led to comparable ground +truth error as the other resampling approaches, but that +systematic resampling outperformed the other approaches in +terms of the overall uncertainty of the filter. Indeed, we note +that multinomial resampling, the most commonly employed +resampling technique, fared quite poorly in terms of the +variance of the filter uncertainty (Fig. 10b and 10d). +B. Baseline Performance Evaluation +Finally, we evaluated the performance of the proposed Shad- +owNav particle filter approach on three test trajectories. Our +analysis consisted of Monte Carlos simulations with 25 seeds +and utilizing 2% odometry noise and initial belief distribution +with σ0 =3 m. Each simulation was run with Ns = 100 +particles and systematic resampling as the resampling scheme +with Neff,thresh = 50 as the resampling threshold. +Figure 11 shows Monte Carlo simulation results for the three +test trajectories. We saw that the initial uncertainty in the filter +began at approximately 3 m as expected by sampling from a +distribution with σ0 =3 m. Thereafter, the filter was able to +improve the rover position estimate, which led to an absolute +error reduction of 4 m. Further, we see in Table 2 that the +metrics computed at the final time step indicate convergence +of the filter, with an average final error of ≤4 m and an +absolute error reduction of 4 m. +As seen in Figure 13, while the filter performed well on +trajectories 2 and 3, the filter was less performant for the +trajectory 1 test case. Figure 12 illustrates the performance +of the filter on trajectory 1 for two different random seeds +as the rover starts from the northern edge of the orbital map +and moves southward. +During the middle portion of this +traverse, the craters were out-of-sight for the rover and, as we +9 + +[x-x| [m] +6 +5 +[u] Ix-> +4 +X +3 +Resampling +2 +Residual +Systematic +1 +Stratified +Multinomial +0 +0 +20 +40 +60 +80 +100 +120 +Iterationg +Resampling +Residual +4 +Systematic +Stratified +Multinomial +3 +[m] +2 +1 +0 +0 +20 +40 +60 +80 +100 +120 +Iteration[x-x] [m] +7 +Resampling +Residual +6 +Systematic +Stratified +5 +Multinomial +[m] +4 +[x-) + +4 +X +3 +2 +1 +0 +0 +20 +40 +60 +80 +100 +120 +140 +Iterationg +4.0 +Traj Name +Traj. 1 +3.5 +Traj. 2 +Traj. 3 +3.0 +2.5 +[w] +2.0 +1.5 +1.0 +0.5 +0.0 +0 +20 +40 +60 +80 +100 +120 +140 +Iteration(a) Trajectory 1 +(b) Trajectory 2 +(c) Trajectory 3 +Figure 13: The final ground truth error distribution for 25 Monte Carlo simulations showed filter convergence to ≤ 4m +error in all cases for trajectories 2 and 3 and for the majority of cases for trajectory 1. +Table 2: The metrics computed at the end of a long-range +lunar traverse indicate convergence of the particular filter +on trajectories 2 and 3, but spurious measurements from +unlabeled crater lead to relatively poor performance on +trajectory 1. +Error +Uncertainty +Mahalanobis Dist. +Traj. 1 +3.84 ± 2.78 +1.84 ± 1.12 +8.74 ± 10.03 +Traj. 2 +1.75 ± 0.78 +1.32 ± 0.76 +2.75 ± 1.88 +Traj. 3 +1.68 ± 0.7 +1.39 ± 0.61 +2.92 ± 1.91 +see in Figure 11, false positive observations led to increases +in the error and uncertainty in the filter. +As the crater in +the southern portion of the orbital map became observable +for the rover, we saw that the estimate quickly improved in +case A (Fig. 12a), but continues to have a residual error in +case B (Fig. 12b). This poor convergence behavior was also +explained by false positive observations, wherein the filter +had difficulty reconciling the front edge of the rim with the +back edge, an issue that requires further investigation. +C. Debugging +When testing the particle filter, we found it helpful to generate +“perfect” datasets where ground truth depth was generated +directly from the simulator as shown in Figure 14b) and crater +edges were plotted into the rover frame using their exact +known world coordinates (see Fig. 8a-8c). +This approach +uncovered bugs with our perception and projection pipeline +as well as the particle filter pipeline and it is highly recom- +mended to build such a dataset for all similar work. +7. CONCLUSIONS +In this work we present a system to perform autonomous +absolute localization on a Lunar rover while it is in darkness. +This system entails using a stereo camera and illuminator. We +enhanced a Blender based simulation with a custom Lunar +texture and an implementation of the Hapke model to model +surface reflectance as accurately as possible. +We further +demonstrate both geometric and learning based techniques +for detecting the leading edge of a crater with ability to +detect some craters out to 20 m range. We propose a method +of matching the detected leading crater rims with known +craters within an orbital map and using these matches to score +(a) +Simulated +image +from +Blender. +(b) Perfect depth: blue is close, +red is far +Figure 14: Crater 1 viewed from 5 m away from front rim. +observations with our Q-Score. Finally we demonstrate abso- +lute localization within our simulation environment with less +than 4 m error, and an absolute error reduction of 4 m upon +detecting craters. +These results show promise for further +investigation in the future on more simulation environments +as well as on to be collected real analogue datasets. +D. Future Work +In the future, we seek to perform several updates and addi- +tional evaluations. The primary focus is to experimentally +collect a nighttime dataset using representative hardware in +an analogue Lunar environment with negative obstacles to +evaluate the system. +Additional evaluation is planned to +evaluate the performance of the proposed approach along +longer trajectories, on more varied Lunar type locales, and +for different rover specific parameters such as camera height +off of the ground. Finally, we plan to validate our proposed +approach on a flight-like embedded computer (e.g., a Snap- +dragon) to demonstrate that it is computationally feasible for +use onboard a Lunar rover. +ACKNOWLEDGMENTS +The research was carried out at the Jet Propulsion Labo- +ratory, California Institute of Technology, under a contract +with the National Aeronautics and Space Administration +(80NM0018D0004). The authors would like to thank Yang +Cheng, Olivier Lamarre, and Scott Tepsuporn for their dis- +cussions during the development of this work. +10 + +14 +12 +10 +Count +8 +6 +4 +2 +0 +0 +2 +4 +6 +8 +10 +Final GT Error [m]14 +12 +10 +Count +8 +6 +4 +2 +0 +0 +2 +4 +6 +8 +10 +Final GT Error [m]14 +12 +10 +Count +8 +6 +4 +2 +0 +0 +2 +4 +6 +8 +10 +Final GT Error [m]REFERENCES +[1] +K. Ennico-Smith, A. Colaprete, R. Elphic, J. Captain, +J. Quinn, and K. Zachny, “The Volatiles Investigating +Polar Exploration Rover payload,” in Lunar and Plane- +tary Science Conference, 2020. +[2] +A. Colaprete, R. C. Elphic, M. Shirley, K. Ennico- +Smith, D. S. S. Lim, Z. Zacny, and J. Captain, +“The Volatiles Investigating Polar Exploration Rover +(VIPER) mission – measurements and constraints,” in +Lunar and Planetary Science Conference, 2021. +[3] +National Academies of Sciences, Engineering, and +Medicine, “Origins, worlds, and life: A decadal strat- +egy for planetary science and astrobiology 2023–2032.” +National Academy Press, Tech. Rep., 2022. +[4] +J. T. Keane, S. M. Tikoo, and J. Elliott, “Endurance: Lu- +nar South Pole-Atken Basin traverse and sample return +rover,” National Academy Press, Tech. Rep., 2022. +[5] +L. Matthies, S. Daftry, S. Tepsuporn, Y. Cheng, D. Atha, +R. M. Swan, S. Ravichandar, and M. Ono, “Lunar +rover localization using craters as landmarks,” in IEEE +Aerospace Conference, 2022. +[6] +S. Daftry, Z. Chen, Y. Cheng, S. Tepsuporn, B. Coltin, +U. Naam, M. M. Lanssie, S. Khattak, M. Deans, +and L. Matthies, “LunarNav: Crater-based localization +for long-range autonomous rover navigation,” in IEEE +Aerospace Conference, 2023. +[7] +H. Hiesinger, C. H. van der Bogert, J. H. Pasckert, +L. Funcke, L. Giacomini, L. R. Ostrach, and M. S. +Robinson, “How old are young lunar craters?” Journal +of Geophysical Research, vol. 117, no. 12, pp. 1–15, +2012. +[8] +M. S. Robinson, +S. M. Brylow, +M. Tschimmel, +D. Humm, S. J. Lawrence, P. C. Thomas, B. W. Denevi, +E. Bowman-Cisneros, J. Zerr, M. A. Ravine, M. A. +Caplinger, F. T. Ghaemi, J. A. Schaffner, M. C. Malin, +P. Mahanti, A. Bartels, J. Anderson, T. N. Tran, E. M. +Eliason, A. S. McEwen, E. Turtle, B. L. Jolliff, and +H. Hiesinger, “Lunar Reconnaissance Orbiter (LROC) +camera instrument overview,” Space Science Reviews, +vol. 150, 2010. +[9] +E. Cisneros, A. Awumah, H. M. Brown, A. C. Mar- +tin, K. N. Paris, R. Z. Povilaitis, A. K. Boyd, M. S. +Robinson, and LROC Team, “Lunar Reconnaissance +Orbiter camera permanently shadowed region imaging +– atlas and controlled mosaics,” in Lunar and Planetary +Science Conference, 2017. +[10] Blender - a 3D modelling and rendering package. Soft- +ware available from https://www.blender.org/. +[11] J. V. Hook, R. Schwartz, K. Ebadi, K. Coble, and +C. Padgett, “Topographical landmarks for ground-level +terrain relative navigation on Mars,” in IEEE Aerospace +Conference, 2022. +[12] K. Ebadi, K. Coble, D. Atha, R. Schwartz, C. Padgett, +and J. V. Hook, “Semantic mapping in unstructured en- +vironments: Toward autonomous localization of plane- +tary robotic explorers,” in IEEE Aerospace Conference, +2022. +[13] J. W. Hwangbo, K. Di, and R. Li, “Integration of orbital +and ground image networks for the automation of rover +localization,” in American Society for Photogrammetry +and Remote Sensing Annual Conference, 2009. +[14] D. Hurley, P. Prem, A. Stickle, C. Hibbitts, A. Deutsch, +A. Colaprete, R. Elphic, S. Li, P. Lucey, Y. Liu, +S. Hosseini, K. D. Retherford, K. Zacny, J. Atkin- +son, M. Benna, W. Farrell, D. Needham, L. Gertsch, +M. Delitsky, and P. Hayne, “Science from the lunar per- +manently shadowed regions,” in Lunar and Planetary +Science Conference, 2020. +[15] B. Wu, R. W. K. Potter, P. Ludivig, A. S. Chung, +and T. Seabrook, “Absolute localization through orbital +maps and surface perspective imagery: A synthetic lu- +nar dataset and neural network approach,” in IEEE/RSJ +Int. Conf. on Intelligent Robots & Systems, 2019. +[16] V. Franchi and E. Ntagiou, “Planetary rover localisa- +tion via surface and orbital image matching,” in IEEE +Aerospace Conference, 2022. +[17] A. Howard, A. I. Ansar, T. E. Litwin, and S. B. Gold- +berg. (2009) Jet Propulsion Laboratory Stereo Vision +software suite (JPLV). Software available from https: +//software.nasa.gov/software/NPO-18593-1T. +[18] H. Hirschmuller, “Stereo processing by semiglobal +matching and mutual information,” IEEE Transactions +on Pattern Analysis & Machine Intelligence, vol. 30, +no. 2, pp. 328–341, 2008. +[19] J. Canny, “A computational approach to edge detection,” +IEEE Transactions on Pattern Analysis & Machine In- +telligence, vol. 8, no. 6, pp. 679–698, 1986. +[20] S. Xie and Z. Tu, “Holistically-nested edge detection,” +in IEEE Int. Conf. on Computer Vision, 2015. +[21] M. S. Arulampalam, S. Maskell, N. Gordon, and +T. Clapp, +“A tutorial on particle filters for on- +line nonlinear/non-Gaussian Bayesian tracking,” IEEE +Transactions on Signal Processing, vol. 50, no. 2, pp. +174–188, 2002. +[22] C. Gentner, S. Zhang, and T. Jost, “Log-PF: Particle +filtering in logarithm domain,” Journal of Electrical and +Computer Engineering, vol. 2018, 2018. +[23] B. Hapke, Theory of reflectance and emittance spec- +troscopy. +Cambridge Univ. Press, 2012. +[24] ——, “Bidirectional reflectance spectroscopy: 5. the +coherent backscatter opposition effect and anisotropic +scattering,” Icarus, vol. 157, no. 2, pp. 523–534, 2002. +[25] F. Schmidt and S. Bourguignon, “Efficiency of BRDF +sampling and bias on the average photometric behav- +ior,” Icarus, vol. 317, pp. 10–26, 2019. +[26] X. Xu, J. Liu, D. Liu, B. Liu, and R. Shu, “Photo- +metric correction of Chang’E-1 interference imaging +spectrometer’s (IIM) limited observing geometries data +with Hapke model,” Remote Sensing, vol. 12, no. 22, p. +3676, 2020. +[27] S. Joshi and S. Boyd, “Sensor selection via convex +optimization,” IEEE Transactions on Signal Processing, +vol. 57, no. 2, 2009. +[28] L. Carlone and S. Karaman, “Attention and anticipation +in fast visual-inertial navigation,” IEEE Transactions on +Robotics, vol. 35, no. 1, 2019. +[29] T. Li, M. Boli´c, and P. M. Djuri´c, “Resampling meth- +ods for particle filtering,” IEEE Transactions on Signal +Processing, vol. 32, no. 3, pp. 70–86, 2015. +[30] C. Kuptametee and N. Aunsri, “A review of resampling +techniques in particle filtering framework,” Measure- +ment, vol. 193, 2022. +11 + +BIOGRAPHY[ +Abhishek Cauligi is a Robotics Tech- +nologist in the Robotic Surface Mobility +Group at NASA Jet Propulsion Labo- +ratory, California Institute of Technol- +ogy. +Abhishek received his B.S. in +Aerospace Engineering from the Uni- +versity of Michigan - Ann Arbor and +his PhD. in Aeronautics and Astronau- +tics from Stanford University. +His re- +search interests lie in leveraging recent +advances in nonlinear optimization, machine learning, and +control theory towards planning and control for complex +robotic systems. +R. Michael Swan is a Robotics Systems +Engineer at NASA Jet Propulsion Lab- +oratory, California Institute of Technol- +ogy. He received his B.S. in Computer +Engineering from Walla Walla Univer- +sity and his M.S. in Computer Science +from the University of Southern Califor- +nia. He is interested in robotic surface +and aerial autonomy, perception, simu- +lation, and robotic system architecture. +Hiro Ono is the Group Leader of +the Robotic Surface Mobility Group at +NASA Jet Propulsion Laboratory, Cal- +ifornia Institute of Technology. +Since +he joined JPL in 2013, he has led a +number of research projects on Mars +rover autonomy, as well as three NIAC +studies on Enceladus Vent Explorer and +Comet Hitchhiker. Hiro was a flight soft- +ware developer of M2020’s Enhanced +AutoNav and the lead of M2020 Landing Site Traversability +Analysis. He also led the development of a machine learning- +based Martian terrain classifier, SPOC (Soil Property and +Object Classification), which won JPL’s Software of the Year +Award in 2020. +Shreyansh Daftry is a Robotics Tech- +nologist at NASA Jet Propulsion Labora- +tory, California Institute of Technology. +He received his M.S. degree in Robotics +from the Robotics Institute, Carnegie +Mellon University, and his B.S. degree +in Electronics and Communication En- +gineering. +His research interest lies +at the intersection of space technology +and autonomous robotic systems, with +an emphasis on machine learning applications to percep- +tion, planning, and decision making. +At JPL, he is the +Group Leader of the Perception Systems group, is working +on the Mars Sample Recovery Helicopter mission, and has +led/contributed to technology development for autonomous +navigation of ground, airborne, and subterranean robots. +John Elliott is a principal engineer +in JPL’s Mission Concept Systems De- +velopment group. +He currently serves +as Program Engineer for the Planetary +Science Formulation office. His recent +tasks have included serving as study +lead for the Planetary Decadal Survey’s +three lunar rover mission concept stud- +ies, Intrepid, INSPIRE, and Endurance, +and performing systems engineering and +leadership roles on a number of recent Discovery and New +Frontiers mission proposals. Mr. Elliott’s past experience +includes six years in the terrestrial nuclear power industry +with Bechtel Corporation in addition to 30 years in aerospace +systems at TRW and JPL. +Larry Matthies is the technology co- +ordinator in the Mars Exploration Pro- +gram Office at JPL. He received B.S., +M. Math,and PhD degrees in Computer +Science from the University of Regina +(1979), University of Waterloo (1981), +and Carnegie Mellon University (1989). +He has been with JPL for more than +32 years. He has conducted technology +development in perception systems for +autonomous navigation of robotic vehicles for land, sea, +air, and space. +He supervised the JPL Computer Vision +group for 21 years. He led development of computer vision +algorithms for Mars rovers, landers, and helicopters. He is a +Fellow of the IEEE and a member of the editorial boards of +Autonomous Robots and the Journal of Field Robotics. +Deegan Atha is a Robotics Technologist +within the Perception Systems Group of +the Mobility and Robotic Systems Sec- +tion at the Jet Propulsion Laboratory. +He received his B.S. degree from Purdue +University in Electrical Engineering and +his M.S. in Computer Science from the +Georgia Institute of Technology. He is +currently the Principal Investigator for +the ShadowNav task and leading the se- +mantic perception effort for the DARPA RACER project. His +interests are in the infusion of machine learning and robotic +perception into autonomous systems operating in unstruc- +tured environments. +12 + diff --git a/29E3T4oBgHgl3EQfoAqh/content/tmp_files/load_file.txt b/29E3T4oBgHgl3EQfoAqh/content/tmp_files/load_file.txt new file mode 100644 index 0000000000000000000000000000000000000000..b3476a345a87a2443b7cb91e416d0fde40549cb5 --- /dev/null +++ b/29E3T4oBgHgl3EQfoAqh/content/tmp_files/load_file.txt @@ -0,0 +1,1161 @@ +filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf,len=1160 +page_content='ShadowNav: Crater-Based Localization for Nighttime and Permanently Shadowed Region Lunar Navigation Abhishek Cauligi* abhishek.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='s.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='cauligi@jpl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='nasa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='gov R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Michael Swan* robert.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='swan@jpl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='nasa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='gov Hiro Ono masahiro.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='ono@jpl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='nasa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='gov Shreyansh Daftry shreyansh.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='daftry@jpl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='nasa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='gov John Elliott john.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='o.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='elliott@jpl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='nasa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='gov Larry Matthies lhm@jpl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='nasa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='gov Deegan Atha deegan.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='j.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='atha@jpl.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='nasa.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='gov Jet Propulsion Laboratory, California Institute of Technology Pasadena, CA 91109, USA Abstract—There has been an increase in interest in missions that drive significantly longer distances per day than what has currently been performed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' For example, Endurance-A proposes driving several kilometers a day in order to reach its target traverse of 2000 km in 4 years.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Additionally, some of these proposed missions, including Endurance-A and rovers for Permanently Shadowed Regions (PSRs) of the moon, re- quire autonomous driving and absolute localization in darkness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Endurance-A proposes to drive 1200 km of its total traverse at night.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The lack of natural light available during these missions limits what can be used as visual landmarks and the range at which landmarks can be observed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In order for planetary rovers to traverse long-ranges, onboard absolute localization is critical to the rover’s ability to maintain its planned trajectory and avoid known hazardous regions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Currently, the localization performed onboard rovers is relative to the rover’s frame of reference and is performed through the integration of wheel and visual odometry and inertial measurements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' To accomplish absolute localization, a “ground-in-the-loop” (GITL) operation is performed wherein a human operator matches local maps or images from onboard with orbital images and maps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This GITL operation places a limit on the distance that can be driven in a day to a few hundred meters, which is the distance that the rover can maintain acceptable localization error via relative methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Previous work has shown that using craters as landmarks is a promising approach for performing absolute localization on the moon during the day.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In this work we present a method of absolute localization that utilizes craters as landmarks and matches detected crater edges on the surface with known craters in orbital maps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' We focus on a localization method based on a perception system which has an external illuminator and a stereo camera.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' While other methods based on lidar exist, lidar is not currently planned for deployment on the current proposed nighttime and PSR missions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In this paper, we evaluate (1) both monocular and stereo based surface crater edge detection techniques, (2) methods of scoring the crater edge matches for optimal localization, and (3) localization performance on simulated Lunar surface imagery at night.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' We demonstrate that this technique shows promise for maintaining absolute localization error of less than 10 m required for most planetary rover missions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' TABLE OF CONTENTS 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' INTRODUCTION.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 1 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' RELATED WORKS .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 2 Abhishek Cauligi and R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Michael Swan contributed equally to this work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 978-1-6654-9032-0/23/$31.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='00 ©2023.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' California Institute of Technology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Government sponsorship acknowledged.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The research was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration (80NM0018D0004).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' APPROACH.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 3 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' DATASETS OVERVIEW .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 5 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' CRATER DETECTION PERFORMANCE .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 6 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' LOCALIZATION PERFORMANCE .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 8 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' CONCLUSIONS .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 10 ACKNOWLEDGMENTS .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 10 REFERENCES .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 11 BIOGRAPHY .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 12 L R Figure 1: The ShadowNav localization algorithm per- forms absolute localization for a Lunar rover mission located at the red position in the left image by matching known craters from (left) an orbital map against (right) detected craters from the rover stereo cameras.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' INTRODUCTION Long-range Lunar navigation, and specifically navigating within darkness, has gained a significant amount of traction recently.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' For example, missions to Permanently Shadowed Regions (PSRs) of the moon have been proposed such as the VIPER mission [1], [2] and the Lunar Polar Volatiles Explorer mission concepts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Furthermore, there are missions that have proposed driving during the Lunar night in order to traverse longer distances.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' For example, the new Decadal Survey [3] recommends the Endurance-A Lunar rover mis- sion should be implemented as a strategic medium-class mission as the highest priority of the Lunar Discovery and Exploration Program.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The Endurance-A rover proposal plans to drive 2000 km in the South Pole-Aitken (SPA) Basin to collect 100 kg of samples, which would be delivered to Artemis astronauts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This mission concept study [4] identified several key capabilities required to complete this mission which are: (1) Endurance will need to drive 70% of its total distance during the night to enable daytime hours dedicated to science and sampling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' (2) The mission will require on- board autonomy for the majority of its operations, while the 1 arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='04630v1 [cs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='RO] 11 Jan 2023 Orbital Image Generation Stereo Images Generation Crater Edge Detection Local-to-Global Transform Particle Filter Step Compute Q-Score Orbital Image Generation Particle Filter Disparity Generation Disparity Hole Filler Figure 2: Schematic of the ShadowNav algorithm proposed to perform absolute localization on the Moon.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' A particle filter is used to match craters detected by the rover stero cameras with known craters from an orbital map.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' ground only handles contingencies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' (3) Global localization is necessary to maintain an error of <10 m relative to orbital maps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' At present, existing rovers perform onboard localization rel- ative to their own reference frame.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This is accomplished by using wheel and visual odometry and inertial measure- ments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Absolute localization is performed periodically with a “ground-in-the-loop” (GITL) operation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This is acceptable for current driving distances which are a few hundred meters a day.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Existing relative localization has around 2% drift and therefore can only drive at most 500 m before the error will be larger than 10 m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In order to traverse longer distances, on the order of several kilometers a day proposed by missions such as Endurance-A, autonomous absolute localization be- comes critical.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' At present the Lunar surface does not have continuous communication with Earth.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Therefore, having to perform several GITL operations for absolute localization in a day will significantly reduce the distance that can be driven.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The lack of frequent absolute localization for the rover would lead to errors greater than the maximum 10 m localization error which would present significant risks to the mission through deviations from the desired trajectory and risk for unidentified obstacles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Craters as landmarks have been shown to be promising for absolute localization on the Moon [5], [6].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' However, the lack of natural light available while driving within a PSR or during the Lunar night limits what can be used as a landmark and the range at which the landmarks can be observed.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Using craters is still promising as the average distance between craters of ≥10 m in diameter is 100 m on terrain with relatively fresh craters and 10 m on terrain with old craters [7].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Addi- tionally the Lunar Reconnaissance Orbiter Camera (LROC) provides digital elevation models (DEMs) with a resolution between 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='5 m-5 m per pixel [8] and there are some DEMs within PSRs [9].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In this work, we propose using a stereo camera with an illuminator positioned below the stereo camera in order to detect crater rims within the darkness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The use of such an illuminator is motivated by the Endurance-A mission concept study [4], which proposes the use of a stereo camera with an illumination source as the perception system for a rover operating in darkness.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Global localization is then accom- plished by matching the detected crater rims against known craters from an orbital image as shown in Figure 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' To handle the uncertainty and nonlinearity of the crater rim detection model, we utilize a particle filter with a novel Q-Score metric for ranking potential crater matches in order to estimate the absolute position of the rover within an orbital map.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This paper demonstrates the initial results of both crater detection within darkness and absolute localization within simulation which are the results of the first two years of a planned three year effort to validate this approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Work is ongoing to collect and validate this approach in a real-world Lunar analogue test location.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Statement of Contributions: This paper presents an approach to absolute localization on the Moon that can be performed while a rover is in darkness, such as within a PSR or during the Lunar night.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The main contributions of the work as summarized below: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' We developed a simulator based on Blender [10] which renders simulated surface stereo imagery of the Lunar sur- face in darkness located within a known orbital position.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The rendering process utilizes the Hapke lighting model for more accurate surface reflectance as well as DEMs captured by LROC for realistic crater distributions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' We evaluated different crater-edge detection techniques and demonstrate a method which captures 80% of the leading crater arc at 10 m and can detect crater arcs out to 20 m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' We present a method to localize a rover within an orbital map using surface crater-edge detections and known orbital craters based on a particle filter and a metric we call the Q- Score which is detailed in Section 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' We demonstrate our absolute localization technique can achieve less than 2 m absolute error with an assumed odome- try drift of 2% and an initial 3-sigma uncertainty of 3 m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' RELATED WORKS Absolute localization on planetary surfaces is critical for expanding the range rovers can travel in a day and over the course of a mission and there have been many previous works that investigate this problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' There have been techniques proposed for the Martian surface.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Works such as [11], [12] consider far range and horizon features which are at ranges that are beyond what is expected can be seen in the dark.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 2 Figure 3: Figure demonstrating the impact of placement of light source on crater rim shadows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Left: Sample render of a crater with light source even with camera.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Right Sample render of a crater with light source below the camera.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' [13] proposes a technique on the Martian surface for absolute localization that uses rocks and DEMs surface features.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In our work, we focus on the problem of global localization in darkness which is relevant for permanently shadowed regions of the moon, for which there has been a surge of interest in conducting scientific measurements and activities [14].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Our solution approach is inspired by a host of recent works that seek to leverage orbital maps for global rover localization in these shadowed regions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In [13], the authors propose a local- ization procedure that matches an observed rover image with an orbital map, but this approach neglects the rover motion model and yields a deterministic estimate of the robot belief.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' A purely data-driven approach is presented in [15], wherein a convolutional neural network is trained on synthetic data to match the rover observations with orbital imagery.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Closest to our approach, [16] presents a particle filtering technique to compare rover monocular camera imagery with orbital imagery and uses a Siamese neural network approach to assign each particle a likelihood weight.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The authors in [6] propose a similar approach for Lunar absolute localization known as LunarNav.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' However, LunarNav focuses on the day- time localization problem and therefore considers different methods of crater matching that rely on greater knowledge of the surface geometry than available in the nighttime case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' APPROACH In this work, we propose an absolute localization approach which utilizes a crater’s leading edge as landmarks for local- ization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The end result of this approach will be an estimated position and uncertainty within the orbital frame.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' At present, this approach only considers position localization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Rover orientation is assumed to be given by a star tracker which can compute orientation in three dimensions from celestial measurements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Our approach consists of two primary com- ponents: 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' A leading-edge crater detection methodology for use with a Lunar rover equipped with a stereo camera system and illumination source.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' A particle filter for computing a position belief based on a score computed based on the association of crater edges and known orbital ground truth craters, which we call the Q- Score, and the robot motion model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Surface Crater Detection In order to identify craters on the surface, the system was de- signed to be used in conjunction with a perception system that contained a stereo camera and an illumination source.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This perception system was configured where the illumination source was beneath the stereo camera.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Examples of simulated images with the light at the same height as the cameras and the light positioned beneath the cameras are in Figure 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' It was observed that placing the illumination source below the camera results in a shadow at the leading edge of a negative obstacle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Furthermore, offsetting the light with the cameras reduced the impact of the Hapke model washing out some of the surface texture.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Further details on the Hapke model and its impact on surface terrain are provided in Section 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Here, we first review the three different techniques studied in this work for detecting a crater’s leading edge: (1) a method of detecting jumps within stero disparities, (2) a Canny edge detector used to find the shadow on the leading edge, and (3) a convolutional neural network (CNN)-based edge detector that uses both the monocular and disparity image as input.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Stereo Disparity Discontinuity Method The first approach for leading edge crater detection relies on detecting discon- tinuities within the stereo disparity image.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' To accomplish this, the stereo disparity image must first be generated using methods such as the JPLV algorithm [17] or the Semi-Global Block Matching (SGBM) approach [18], among others.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' To account for the low contrast that may be present in the Lunar rover case, Contrast Limited Adaptive Histogram Equaliza- tion (CLAHE) is first run on the input images prior to running stereo.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' CLAHE is an adaptive histogram equalization and operates on sub-regions of an image which allows more consistent equalization across different lighting conditions within an image.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This is useful for this application as there is a light-to-dark gradient from near-to-far within the images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The resulting disparity image is then scanned column-by- column and, when the difference between any two disparities is greater than some pre-defined threshold, the larger column index is marked as a crater edge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Further, any numerical issues stemming from stereo holes are accounted for by omitting any pixels with spurious values during comparison.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Canny Edge Detector Method For sensor configura- tions that contain an illuminator located beneath the stereo cameras, shadows appear on the leading edge of negative obstacles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In such cases, a Canny edge detector can be used to distinguish the stark contrasting dark line along the rim.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In this work, the Canny edge detector from OpenCV [19] is used to find these shadows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' CNN-Based Edge Detector Method The Holistically- Nested Edge Detection (HED) approach presents a CNN- based deep learning based method for leading edge crater detection [20].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This method uses the HED approach and can be performed by directly using the publicly released neural network weights.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' HED is capable of performing both monocular and stereo depth based edge detection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' For HED to perform edge detection within a depth image, it generates a three channel image that contains horizontal disparity, height above ground, and angle of the local surface normal with the inferred direction of gravity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The RGB and depth predictions of the CNN are then merged to generate the desired output.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Positive Obstacle False Positive Rejection—One shortcoming of the aforementioned leading edge crater detection approach is the susceptibility of false positive cases in the presence of positive obstacles.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In order to account for this positive 3 Algorithm 1 Q-Score Computation Require: Belief bt i, set of crater observations {zt 0,rover, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', zt m,rover}, set of ground truth craters {ct 0,world, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', ct ℓ,world}, positive value ε 1: Qinc ← ε 2: for i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' , m do 3: zt 0,world ← rover to world(zt 0,rover) 4: dcr ← min ∥cj,world − zt 0,world∥ 5: Qinc ← Qinc + dcr 6: end for 7: Qscore ← min � 1, ( 1 mQinc)−1� 8: return Qscore Algorithm 2 ShadowNav Particle Filtering Algorithm Require: Initial belief distribution (µ0, Σ0), number of particles Ns, number of effective particles threshold Neff,thresh 1: {b0 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', b0 Ns} ← sample beliefs(µ0, Σ0) 2: {w0 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', w0 Ns} ← {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', 1} 3: t ← 1 4: while particle filter running do 5: {zt 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', zt m} ← get observations() 6: {qt 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', qt Ns} ← {0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', 0} 7: for i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', Ns do 8: bt i ← propagate sample(bt−1 i ) 9: qt i ← log Q score(bt i, {zt 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', zt m}) 10: end for 11: qt min ← min(qt 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', qt Ns) 12: for i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', Ns do 13: wt i ← wt−1 i + qt i − qt min 14: end for 15: Neff ← compute Neff(wt 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', wt Ns) 16: if Neff ≤ Neff,thresh then 17: {bt 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', bt Ns} ← resample beliefs({bt i}Ns i=1, {wt i}Ns i=1) 18: {wt 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', wt Ns} ← {1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', 1} 19: end if 20: t ← t + 1 21: end while obstacle issue, the detected edge points are passed through a filter that removes points which have hits on the far side of the crater edge with a detected negative or flat slope.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Detected edge points are kept only if, within the region directly beyond the detected edge, there exists a positive slope or if there is not enough stereo to accurately compute the slope.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Thus, the case of a detected positive slope is assumed to correspond to the rising edge of the crater under the assumption that the detected edge is the leading edge of a negative obstacle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Alternatively, a detected edge is also retained if the far edge is not captured due to low light conditions, as this is assumed to be an indication of the presence of a large crater.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Particle Filter Here, we provide an overview of the proposed ShadowNav particle filtering approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' First, we provide further details on the Q-Score metric that is used in the belief update step.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Q-Score—The Q-Score provided the measurement probabil- ity of some position belief based on rover frame observations and an orbital map.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The procedure for computing the Q-Score is given in Algorithm 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The algorithm takes as input a given belief bt i, a set of m observed edges in rover frame, and a set of ℓ ground truth crater observations to associate these mea- surements with (Line 1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' A value Qinc is initialized to some negligibly small, positive value ε to later avoid divide-by-zero Algorithm 3 Systematic Resampling Require: Particles {bt 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', bt Ns} and associated weights {wt 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', wt Ns} 1: nt = log � �Ns i=1 exp(bt i) � 2: { ˜wt 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', ˜wt Ns} ← {0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', 0} 3: for i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', Ns do 4: ˜wt i ← exp(wt i − nt) 5: end for 6: {q0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', qNs} ← cum sum({ ˜wt 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', ˜wt Ns}) 7: n ← 0 8: m ← 0 9: u0 ∼ U(0, 1 Ns ) 10: while n ≤ Ns do 11: u = u0 + n Ns 12: while qm ≤ u do 13: m ← m + 1 14: end while 15: n ← n + 1 16: bt n ← bt m 17: end while 18: return {bt 0, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='..' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', bt Ns} issues (Line 1).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Next, for each measurement zt i in the rover frame, the detected edge is converted to world frame (Line 3) and the minimum distance to an edge from the ground truth map computed (Line 4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The Qinc is incremented by the distance between the observed edge and its associated ground truth observation (Line 5).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The Q-Score is computed as the reciprocal of Qinc and a min operation is applied to ensure that the score provided by any particular run is between 0 and 1 (Line 7).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This implies that observations and belief pairs which are less than 1 m away from ground truth will receive the same score as those exactly 1m away from ground truth, which is seen as acceptable given the orbital DEM resolution and mission concept localization requirements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In addition to the shortest distance formulation from Line 4, additional approaches were also explored for determining the Q-Score.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' One alternate approach investigated included fitting a Gaussian normal distribution on the orbital map crater edges and the Q-Score value was them computed based on the intensity (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', distance to the computed mean) of the point hit by observations or 0 in cases when no point was hit.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In practice, it was determined that the shortest distance formulation provided the most robust results for use with the particle filter and also did not require additional projection calculations to project each belief from the orbital frame to rover frame.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Overview—A description of the ShadowNav particle filtering algorithm is given in Alg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The algorithm takes as input a Gaussian belief distribution (µ0, Σ0) assumed for the initial robot position, the number of particles Ns to use in the particle filter, and a threshold for the effective number of beliefs Neff,thresh used to trigger resampling (Line 2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The filter is initialized by sampling Ns particles from the initial belief distribution and assigning a weight of equal importance for each particle (Lines 1-2).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' As common in particle filtering implementations [21], we note that we used the log of the weights for improved numerical stability of the weight update step [22].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Given a new set of crater observations (Line 5), a set of Q-Score measurements is initialized for computing for each individual particle (Line 6).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' After applying the motion model update to each particle (Line 8), the Q-Score for each updated particle is computed using the procedure from Alg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 1 by comparing against the current measurements 4 (a) Sample of terrain with 90◦ from cam- era.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' (b) Sample opposi- tion effect during the day.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' (c) Sample effect of surface reflectance at night with an illumi- nator.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Figure 4: The opposition effect simulated during the day and its effect at night with an external illuminator.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' (Line 9).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The particle weights are then updated in log-domain (Line 13) with a normalization step to ensure non-negative weights (Line 11).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Next, the number of effective samples Neff at the current iteration is calculated (Line 15).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' A common pitfall of particle filters is “degeneracy”, wherein the weights {wt i} collapse around a handful of particles and computa- tional resources are wasted on propagating low likelihood particles [21].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' If Neff is below the threshold Neff,thresh, then this indicates that the filter is degenerating and a resample operation is triggered (Line 17).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Further details on the systematic resampling approach used in this work are provided in Algorithm 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Given a set of particles and their associated weights, the weights are first normalized to (0, 1] from log-domain (Lines 1-4) and the cumulative sum of these normalized weights ˜wt i computed (Line 6).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The key step in systematic resampling is to sample a random value u0 from a uniform distribution inversely proportional to Ns (Line 9) and then incrementally sample a new particle from this “bin” of width 1 Ns .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This ensures that, after resampling, at least one particle is retained from each 1 Ns interval from the previous belief distribution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Surface to Orbital Crater Transformation For every observation step, rover frame crater edges were detected with a stereo camera pair that provided the depth, and thus a relative position for the crater edge was saved.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This relative crater distance was added to each particle’s belief position to form an estimate of the observed crater position in the world frame for each particle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The orbital map was projected to the world frame and then the shortest distance metric noted in the Q-Score algorithm was used to determine which particle belief positions were most likely and thus which observed crater was the most likely one to match the known orbital craters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Stereo hole filling—As some crater edge detections do not rely on depth information, not all pixels in the stero camera depth or disparity image will have a detected depth value and, in such cases, no relative position would be available for matching rover observations to the orbital map.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' For such observations, a simple plane fit can be carried out to fill in the depth information.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' A future area of investigation includes carrying out an improved stereo hole filling approach, in particular using existing knowledge on what the regional terrain looks like.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Figure 5: The trajectories used for the numerical ex- periments are overlaid on the orbital map here with the crater numbers in black.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' A red square indicator is at the start and a green circle indicator is at the end of each trajectory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Trajectory 1 is in blue, trajectory 2 is in orange, and trajectory 3 is in purple.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' DATASETS OVERVIEW A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Simulated Lunar Environment At the time of writing, a Lunar dataset with images captured in the dark with an illuminator did not exist.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Therefore, to evaluate the approach, a simulation environment was devel- oped using the Blender software [10].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In order to simulate images as realistically as possible, the Hapke lighting model [23], [24], [25] was implemented.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This model approximates the Lunar surface reflectance and will simulate the “opposi- tion effect”.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This effect leads to a focused point of extreme saturation at a location within an image where the camera ray and light source are at zero phase angle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The Hapke lighting model was implemented using the “old highland” parameters of the moon provided in [26], as these most closely match the poles of the moon where PSRs can be found.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The coherent backscattering opposition effect (CBOE) was left out of our implementation and only the shadow hiding opposition effect (SHOE) was implemented as it dominates most or all lighting calculations in our use case, while CBOE has a negligible or very small effect.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Initial implementation was done using the Open Shading Language (OSL), however not all rays are available for calculation due to optimizations made in OSL, so workarounds were needed to implement the Hapke light- ing model in Blender using OSL.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' While this was partially successful, it was not very robust and we had numerous issues.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Instead of using OSL, we opted to modify the source of Blender to add the Hapke bidirectional reflectance distribution function (BRDF) directly into the Blender Cycles renderer code which also reduced the render time by greater than a factor of 2 through the use of Nvidia CUDA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In order to represent a realistic 3D model of the sur- face geometry, DEMs produced from LROC were utilized.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' While LROC has enough resolution to resolve craters of around 10 m, its resolution is not quite good enough for generating smooth surface imagery.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In order to have smooth surface image renders, the DEMs from LROC were scaled down to be 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='25 m resolution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Crater measurements in fu- 5 10Table 1: Table of crater sizes in crater detection dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Crater Diameter (m) Depth (m) 1 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='2 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='0 2 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='75 3 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='84 4 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='55 5 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='7 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='40 6 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='27 7 11.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='9 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='44 8 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='9 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='48 9 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='1 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='49 10 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='3 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='25 ture discussions were based on this scaled resolution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This scaled DEM was imported into Blender and a surface texture was added.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The surface texture comprised of two scales of fractal Brownian motion, which is a natural noise that was added to the DEM in order to simulate Lunar surface texture for stereo to utilize.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Figure 4 demonstrates three sample renders, two in the daylight and one at night with an illumination source from our simulation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' It demonstrates what the surface looks like in daytime conditions as well as the effect of the Hapke model during the day with the sun behind the camera and the effect of the illumination source.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' From this it was observed that the full amount of daytime texture is not observed during the night with an illumination source.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Simulated Craters for Detection Analysis In order to evaluate the performance of different crater de- tection techniques, a dataset with different sized craters was built.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This dataset was built using the simulation process within Blender and captured stereo pair renders between 5 m and 20 m from the front crater rim in increments of 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='1 m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This dataset contained 10 different craters with varying sizes and depths.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The sizes of the craters within this dataset are in Table 1 and their locations corresponding to the crater ID in our simulated environment are marked in Figure 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Simulated Trajectories for Localization Analysis In order to evaluate the localization performance, several trajectories were run in the simulated environment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' These trajectories were run to generate an image every 1 m and were designed to approach craters in different ways that might present challenges to our filtering approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The 1 m observation delta was used to reduce render times of our dataset, as rendering every 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='1 m did not result in a significant localization performance change.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' An overview of the trajectories within the orbital environment are displayed in Figure 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Real Data of Negative Obstacles at Night In addition to the simulated data generated, a dataset was collected in the Arroyo, which is a dry river bed near the NASA Jet Propulsion Laboratory.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This dataset contained a few different negative obstacles that were imaged at 5, 10, and 15 m away from the leading edge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This dataset was used to validated that the stereo and crater edge detection algorithms work on real data collected at night with an external illuminator.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Figure 6: Plots of different metrics evaluating crater detection performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Left: Plot that shows image-based crater edge detection score versus range for all craters evaluated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Right: Plot that shows percent of the crater front arc detected for all craters evaluated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Figure 7: Sample stereo results using JPLV stereo on a sample negative obstacle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' CRATER DETECTION PERFORMANCE A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Metrics In order to evaluate the performance of surface crater detec- tion, the dataset referenced in Section 4 was utilized.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Five different combinations of algorithms were evaluated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' These were disparity discontinuity detection within SGBM stereo, disparity discontinuity detection within JPLV stereo, HED using SGBM stereo, HED using JPLV stereo, and a hybrid JPLV disparity discontinuity detection and canny edge detec- tion approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The hybrid discontinuity detection and canny approach was implemented so that Canny only ran on the portion of the image that was 10 m away or further.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This was done since it was observed the discontinuity detection worked well in the near range but stereo began to degrade beyond 10 m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' These algorithms were evaluated with two different metrics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The first was an image based edge scoring method which captures an average Gaussian probability that a detected edge is on a ground truth crater edge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' It utilizes a distance error computed in image space as represented in Equation 1 where Errordistpx is the pixel error from ground truth to detection, rangegt is the known ground truth range, fl is the focal length of camera, ss is the sensor size of the camera, and Errordist is the error in meters of the detection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Errordist = Errordistpx ∗ rangegt (fl ∗ ss) (1) 6 5m 10mImageBasedCraterEdgeScoreversusRange 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='0 + Disparity JPLV Disparity SGBM + HED JPLV 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='8 HEDSGBM Disparity + Canny jPLV 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='6 Score + + + 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='4 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='2 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='0 6 8 10 12 14 16 18 20 GT Range (m)Percent Crater Front Arc Detected vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='Range ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='100 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='+ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='Disparity JPLV ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='Percent of Crater Front Arc Detected (%) ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='HED JPLV ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='Disparity SGBM ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='80 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='HED SGBM ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='Disparity JPLV + Canny ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='60 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='++ ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='40 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='0 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='6 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='8 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='10 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='12 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='14 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='16 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='18 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='20 ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='Range(m)(a) Ground Truth at 7 m ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='(b) Ground Truth at 12 m ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='(c) Ground Truth at 17 m ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='(d) JPLV Disparity + Canny at 7 m ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='(e) JPLV Disparity + Canny at 12 m ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='(f) JPLV Disparity + Canny at 17 m ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='(g) JPLV HED at 7 m ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='(h) JPLV HED at 12 m ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='(i) JPLV HED at 17 m ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='Figure 8: The efficacy of the JPLV HED approach over JPLV Disparity + Canny is demonstrated in simulations of ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='crater rim detection overlay samples for crater 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The distance error was then passed into a Gaussian.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The Gaussian probabilities for all of the detected pixels were summed together and normalized by number of detected points to obtain a score.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This scoring method infused ground truth range values to remove the impact of stereo holes and stereo range uncertainty on the projection in order to better isolate the specific performance of the crater detection algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The sigma value for the Gaussian that was used in these experiments was 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='25m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This was chosen because the resolution of the DEM utilized was 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='25 m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Therefore most detections should fall within this boundary if they are highly accurate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The second metric used was ”percent of front arc detected”.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In this metric, there is a ground truth circle of the orbital crater.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Depending on the pose of the simulated cameras, the half arc of the ground truth circle that was nearest the simulated camera was projected into image space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The crater detection was then matched to the half arc and the percentage of the half arc that was successfully identified was determined.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This metric removes the Gaussian component from the first metric;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' however, it does not capture 7 (a) Negative obstacle 5 m away (b) Negative obstacle 10 m away (c) Negative obstacle 10 m away (d) Negative obstacle 5 m away (e) Negative obstacle 15 m away (f) Negative obstacle 10 m away Figure 9: Qualitative edge detections using JPLV disparity discontinuity detection and Canny hybrid on negative obstacles on a real dataset collected in a dry river bed at night.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' These results demonstrate the transferability of the crater detection algorithms from simulated data to a real environment.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' false positives like the first metric.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Detection Results on Simulated Data The results of running the different algorithms on the simu- lated dataset are observed in Figure 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' There were several notable observations from the results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The first was that the algorithms tended to perform the best around 10 m and did not improve as craters came closer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This was believed to be because as the camera gets closer to the crater, more of the crater becomes visible and the discontinuities become smaller.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' However, as the crater becomes further than 10 m, the stereo began to degrade.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Additionally, for the hybrid stereo and Canny technique, the Canny detection started de- tection at 10 m and led to a significant jump in performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In terms of algorithm comparison, JPLV disparity disconti- nuity performed better than SGBM disparity discontinuity which is likely due JPLV having more holes than SGBM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' These holes at the boundary helped the disparity discontinuity detector find a better edge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' However, for HED, it performed well with either stereo technique, likely due to its representa- tion of depth containing height values.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' HED was used with its out of the box weights from its authors.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' It likely could be improved with finetuning on a Lunar dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' In addition to quantitative results, samples of crater rim de- tection overlays are in Figure 8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' These results were on crater 1 which is a nearly 10 m in diameter crater.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Both methods were able to detect the craters well, but JPLV HED did have more falloff at 17 m than the Canny detector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' However, the Canny edge detector was optimized for this environment where as HED was a generalized detector.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Overall the generalization of HED was extremely promising as a crater rim detection approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Detection Results on Real Data As described previously, data was collected from a location with negative obstacles at night.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' This dataset was used to validate the performance of stereo and crater detection algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Figure 7 presents a sample of 5 m and 10 m negative obstacles and the corresponding stereo results from JPLV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' From this figure is was observed that stereo is dense up unto the leading edge of the negative obstacle.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Additionally, at 5 m, the far edge of the negative obstacle was captured in the disparity values.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' At 10 m, the far edge, did contain some disparity values but it was sparse.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' While not fully representative of the Lunar surface, this demonstrated that current stereo techniques do have the capability to work in low light conditions at the ranges necessary.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The data was also used to evaluate the edge detection techniques.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The JPLV disparity discontinuity and Canny edge detection hybrid was found to be the best on simulation data and therefore it was used on the real data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Figure 9 demonstrates sample detections at different ranges.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' These detection results did contain false positives on some of the vegetation as the false positive rejection was not run.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Vegetation is not present on the moon, however, objects such as rocks could present similar issues.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Overall, the negative obstacle edge detection qualitatively performs well.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' LOCALIZATION PERFORMANCE In this section, we provide Monte Carlo results on the per- formance of the proposed ShadowNav filtering algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' For each simulation, we analyzed the performance of the ShadowNav filter on the basis of the following metrics: 8 (a) Ground truth error for traj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' (b) Filter covariance for traj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' (c) Ground truth error for traj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' (d) Filter covariance for traj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Figure 10: A comparison of the four proposed resampling schemes demonstrated that systematic resampling empirically outperforms the other scheme in terms of relatively lower ground truth error and reduced uncertainty in the filter.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' (a) Ground truth error.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' (b) Filter covariance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Figure 11: Monte Carlo simulations for trajectories 1–3 demonstrated the efficacy of the Q-Score based particle filtering approach at accomplishing global rover localiza- tion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' (a) Traj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 1 traverse – case A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' (b) Traj.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 1 traverse – case B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Figure 12: Two Monte Carlo trials for trajectory 1 are illustrated with the ground truth in red and the weighted average belief µt in blue.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The comparatively better per- formance of the filter in case A (left) was due to false positive crater rim measurements in case B (right) that led to worse localization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Ground truth error: We computed the weighted average mean µt = �Ns i=1 wt ibt i at time t for the filter using the particle weights and beliefs and compute the ℓ2-distance to the ground truth gtt, i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=', ∥µt − gtt∥2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Particle filter uncertainty: To capture the uncertainty asso- ciated with the current belief, we additionally computed the weighted covariance matrix Σt = �Ns i=1 ˜wt i(bt i−µt)(bt i−µt), where ˜wt i are the normalized weights detailed in Alg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' The metric we report at each time step was the square root of the largest eigenvalue � λmax(Σt), which corresponded to the worst case variance of the estimation error [27], [28].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Mahalanobis distance: The final metric we computed was the Mahalanobis distance, which measures the distance between and the particle filter distribution and ground truth posi- tion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' We approximately computed this by fitting a Gaussian distribution N(µt, Σt) to the particle filter distribution, for which the Mahalnobis distance is simply a weighted ℓ2-norm � (µt − gtt)T (Σt)−1(µt − gtt).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Resampling Scheme Comparison In this section, we compared the baseline systematic resam- pling approach detailed in Alg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 3 against three other resam- pling methods utilized: multinomial, residual, and stratified (we refer the reader to [21], [29], [30] for a thorough review of these approaches.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=') Figure 10 presents the ground truth error and filter uncertainty for the four different resampling approaches.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' We saw that, for the two trajectories compared in Figure 10, systematic resampling led to comparable ground truth error as the other resampling approaches, but that systematic resampling outperformed the other approaches in terms of the overall uncertainty of the filter.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Indeed, we note that multinomial resampling, the most commonly employed resampling technique, fared quite poorly in terms of the variance of the filter uncertainty (Fig.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' 10b and 10d).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Baseline Performance Evaluation Finally, we evaluated the performance of the proposed Shad- owNav particle filter approach on three test trajectories.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Our analysis consisted of Monte Carlos simulations with 25 seeds and utilizing 2% odometry noise and initial belief distribution with σ0 =3 m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Each simulation was run with Ns = 100 particles and systematic resampling as the resampling scheme with Neff,thresh = 50 as the resampling threshold.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Figure 11 shows Monte Carlo simulation results for the three test trajectories.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' We saw that the initial uncertainty in the filter began at approximately 3 m as expected by sampling from a distribution with σ0 =3 m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Thereafter, the filter was able to improve the rover position estimate, which led to an absolute error reduction of 4 m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Further, we see in Table 2 that the metrics computed at the final time step indicate convergence of the filter, with an average final error of ≤4 m and an absolute error reduction of 4 m.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' As seen in Figure 13, while the filter performed well on trajectories 2 and 3, the filter was less performant for the trajectory 1 test case.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' Figure 12 illustrates the performance of the filter on trajectory 1 for two different random seeds as the rover starts from the northern edge of the orbital map and moves southward.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' During the middle portion of this traverse,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' the craters were out-of-sight for the rover and,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/29E3T4oBgHgl3EQfoAqh/content/2301.04630v1.pdf'} +page_content=' as we 9 [x-x| [m] 6 5 [u] Ix-> 4 X 3 Resampling 2 Residual Systematic 1 Stratified Multinomial 0 0 20 40 60 80 100 120 Iterationg Resampling Residual 4 Systematic Stratified Multinomial 3 [m] 2 1 0 0 20 40 60 80 100 120 Iteration[x-x] [m] 7 Resampling Residual 6 Systematic Stratified 5 Multinomial [m] 4 [x-)