{"version":1,"issue":{"week":"2026-W43","title":"Benchmarks for free-floating capture, lunar legged locomotion, and onboard language models","intro":"A strong week for open tools: a capture benchmark built on ISS hardware and an open lunar rover simulator with shadow ground truth. Alongside them, several takes on legged mobility in lunar gravity.","curator":"Space Robots","publishedAt":"2026-10-26","url":"https://space-robots.org/papers/top/2026-W43/","picks":[{"rank":1,"note":"An open benchmark for capturing free-floating objects with a fixed-base arm. Its tasks are modeled on MELFI sample storage and Robotic Refueling Mission servicing hardware.","paper":{"id":"2610.31064","title":"AstroBench: Benchmarking Free-Floating Capture and Space Robotic Manipulation","authors":["Zizhe Wang","Jia Deng","Xin Lu","Zehao Jia","Yi Huang","Yao Zhu Astroup Lab","Beijing","China","Harbin Institute of Technology","Beijing Research Institute)"],"abstract":"Capturing a loose object inside a spacecraft requires a robot to approach a target that can translate and rotate, and whose motion changes when contact is made. These conditions motivate evaluation beyond manipulation of supported objects. We introduce AstroBench, an open benchmark for free-floating capture and space robotic manipulation with a fixed-base arm. The task suite combines four initial-motion tracks with sample-storage operations based on MELFI hardware and cap and captive-screw servicing based on the Robotic Refueling Mission. The two workcells provide nine task definitions and 29 task-target combinations. A metadata-filtered Objaverse inventory supplies 8,414 object instances for asset expansion, alongside five cabin visual contexts and versioned task instructions. We report two acquisition-focused experiments using existing checkpoints. A frozen $\\pi_{0.5}$ DROID joint-position policy captures 2 of 200 static targets and 4 of 200 moving targets across twenty object identities. A cube-specific ACT policy captures 8 of 10 static targets and 6 of 10 moving targets in a separate experiment. Each experiment retains its recorded observation, action and capture configuration. The results document difficult direct transfer for the tested generalist checkpoint and successful task-specific capture behavior, while identifying concrete settings for studying motion-aware perception and contact-sensitive control. Code, task definitions and editable workcell assets are available at https://github.com/AstroUpLab/AstroBench.","published":"2026-10-22","addedAt":"2026-10-22","year":2026,"venue":{"id":"preprint","label":"Preprint","year":null,"isWorkshop":false,"raw":null},"intersection":"robotics","domains":[{"id":"orbital-robotics","label":"ORBITAL ROBOTICS"},{"id":"in-space-servicing","label":"IN-SPACE SERVICING"}],"techniques":[{"id":"manipulation","code":"T1","label":"Manipulation & Grasping","source":"official","via":["openalex-topic:Robot Manipulation and Learning","title:manipulat","abstract:capturing","abstract:manipulat"]},{"id":"simulation-testbeds","code":"T13","label":"Simulation, Digital Twins & Benchmarks","source":"official","via":["openalex-keyword:benchmark","title:benchmark"]},{"id":"learning","code":"T8","label":"Learning-based Methods","source":"official","via":["openalex-topic:Reinforcement Learning in Robotics"]}],"missions":[],"indexedAs":{"arxiv":["cs.RO"],"openalexTopics":[{"id":"T11701","name":"Space Satellite Systems and Control","score":0.944},{"id":"T10653","name":"Robot Manipulation and Learning","score":0.05},{"id":"T10462","name":"Reinforcement Learning in Robotics","score":0.003}],"ntrsSubjects":[],"ieeeTerms":[]},"cardTag":"RESEARCH // ORBITAL ROBOTICS","citations":0,"links":{"page":"https://space-robots.org/papers/paper/2610.31064","arxiv":"https://arxiv.org/abs/2610.31064","pdf":"https://arxiv.org/pdf/2610.31064","html":"https://arxiv.org/html/2610.31064","doi":null,"code":"https://github.com/AstroUpLab/AstroBench","project":"https://astroup.com/astrobench/"}}},{"rank":2,"note":"MARE: an open-source lunar rover simulator with ground truth for long polar shadows, aimed at perception and navigation development under extreme illumination.","paper":{"id":"2610.28465","title":"MARE: An Open-Source Lunar Rover Simulator with Shadow Ground Truth for Perception and Autonomy","authors":["Raúl Castilla-Arquillo","Miguel Olivares-Mendez"],"abstract":"Lunar surface exploration increasingly relies on autonomous rovers that must perceive and traverse unstructured terrain under extreme illumination conditions. Near the lunar poles, low sun elevations produce long, high-contrast shadows that challenge stereo vision, terrain segmentation, and localization. Developing perception and navigation algorithms for these conditions requires large amounts of sensor data with accurate ground truth, which field campaigns and analog facilities can only provide at high cost and with limited variety. This paper presents MARE (Modular Autonomy and Rendering Environment), an open-source simulation suite for lunar rover perception and autonomy built on Unreal Engine 5 and ROS 2. MARE consists of three integrated components. Terrain Studio generates cratered and rocky lunar Digital Elevation Maps (DEMs) to build simulation scenes. A dedicated Unreal Engine project hosts two physics-based rovers with plugins for stereo, lidar, event, and Single-Photon Avalanche Diode (SPAD) cameras, streaming time-stamped sensor data and pixel-accurate ground truth. Lastly, a Guidance, Navigation, and Control (GNC) stack provides web-based mission control. Validation shows a mean Intersection over Union (IoU) of 0.79 between predicted and rendered shadows at low sun elevations, and stereo depth analysis verifies decreased pixel coverage in shadowed regions relative to sunlit terrain.","published":"2026-10-20","addedAt":"2026-10-20","year":2026,"venue":{"id":"preprint","label":"Preprint","year":null,"isWorkshop":false,"raw":null},"intersection":"both","domains":[{"id":"planetary-robotics","label":"PLANETARY ROBOTICS"}],"techniques":[{"id":"perception","code":"T5","label":"Perception & State Estimation","source":"official","via":["openalex-topic:Robotics and Sensor-Based Localization","openalex-keyword:stereo vision","openalex-keyword:terrain segmentation","openalex-keyword:rover localization"]},{"id":"surface-mobility","code":"T2","label":"Surface Mobility & Terramechanics","source":"official","via":["openalex-topic:Soil Mechanics and Vehicle Dynamics"]},{"id":"gnc","code":"T7","label":"Guidance, Dynamics & Control","source":"official","via":["openalex-keyword:guidance, navigation and control"]}],"missions":[],"indexedAs":{"arxiv":["cs.RO"],"openalexTopics":[{"id":"T10191","name":"Robotics and Sensor-Based Localization","score":0.923},{"id":"T10462","name":"Reinforcement Learning in Robotics","score":0.02},{"id":"T11978","name":"Soil Mechanics and Vehicle Dynamics","score":0.016}],"ntrsSubjects":[],"ieeeTerms":[]},"cardTag":"RESEARCH // TESTBEDS & SIMULATION","citations":0,"links":{"page":"https://space-robots.org/papers/paper/2610.28465","arxiv":"https://arxiv.org/abs/2610.28465","pdf":"https://arxiv.org/pdf/2610.28465","html":"https://arxiv.org/html/2610.28465","doi":null,"code":"https://github.com/raulcastar/MARE-Lunar-Simulator","project":"https://youtu.be/EI4HgbMFbBE"}}},{"rank":3,"note":"Treats low-bit compression of an onboard language model as a ground-segment step. Each module is assigned to a quantizer so the static flight artifact stays behaviorally close to the full-precision model.","paper":{"id":"2610.29302","title":"Compressing Onboard Language Models for Space Robots under a Behavioral-Fidelity Requirement","authors":["Jia Deng","Zizhe Wang","Xin Lu","Yi Huang","Zehao Jia"],"abstract":"Language models are being explored as the interface through which space robots interpret operator commands and select among verified skills. Running them onboard calls for low-bit compression, and flight use adds requirements that generic post-training quantization leaves implicit: a fixed memory budget and storage format, a static artifact under configuration control, and behavior close to the full-precision model validated on the ground. We treat compression as a ground-segment step and study QSelect, which composes one static low-bit model from compatible AWQ, GPTQ, and AutoRound checkpoints by assigning each linear module to one quantizer family and scoring every substitution at the final next-token distribution. The default fidelity mode needs no labels or tuned thresholds and accepts a substitution only when the forward KL divergence from the BF16 reference decreases; an optional target-adapted mode adds answer likelihood on a small labeled set. On DeepSeek-R1-Distill-Qwen-7B, Llama-3.1-8B, and Llama-3.2-3B at 3 and 4 bits, composed models reach lower mean held-out perplexity than every single-family model in all six settings, and the fidelity mode lowers perplexity and KL further. In a matched scaffold, final-output scoring gives lower held-out KL than representation-similarity scoring in all 12 paired runs. With labels from training or development splits of the evaluated task families, the target-adapted mode raises 3-bit pooled accuracy on the two Llama models by 4.3 and 4.5 points over the best single-family model. At 3 bits, unquantized embedding and output layers make up 43-46% of the packed artifact. All evaluations use general-domain benchmarks on a ground GPU; space-domain tasks and flight hardware are left to future work.","published":"2026-10-21","addedAt":"2026-10-21","year":2026,"venue":{"id":"preprint","label":"Preprint","year":null,"isWorkshop":false,"raw":null},"intersection":"both","domains":[],"techniques":[{"id":"foundation-models","code":"T9","label":"Foundation Models & LLMs","source":"official","via":["openalex-topic:Topic Modeling","openalex-keyword:language model compression","openalex-keyword:GPTQ","title:language model"]},{"id":"learning","code":"T8","label":"Learning-based Methods","source":"official","via":["arxiv:cs.LG","openalex-topic:Advanced Neural Network Applications"]}],"missions":[],"indexedAs":{"arxiv":["cs.RO","cs.LG"],"openalexTopics":[{"id":"T10036","name":"Advanced Neural Network Applications","score":0.913},{"id":"T10028","name":"Topic Modeling","score":0.011},{"id":"T11689","name":"Adversarial Robustness in Machine Learning","score":0.007}],"ntrsSubjects":[],"ieeeTerms":[]},"cardTag":"RESEARCH // SPACE AI & AUTONOMY","citations":0,"links":{"page":"https://space-robots.org/papers/paper/2610.29302","arxiv":"https://arxiv.org/abs/2610.29302","pdf":"https://arxiv.org/pdf/2610.29302","html":"https://arxiv.org/html/2610.29302","doi":null,"code":null,"project":null}}},{"rank":4,"note":"Twelve training strategies for a simulated Unitree G1 compared across gravity, terrain and granular soil. Earth-trained controllers keep falling on the Moon, and Moon-trained ones fail on Earth.","paper":{"id":"2610.30080","title":"Do Earth-Trained Humanoid Controllers Walk on the Moon? Gravity, Terrain, and Soil Transfer","authors":["Guillem Casadesus Vila","Sri Aditya Deevi","Grace X. Gao"],"abstract":"Humanoid robots could aid lunar operations, but how to train their locomotion controllers for lunar gravity, terrain, and soil is unclear. We compare 12 strategies for a simulated Unitree G1 across gravity, terrain, and rigid or granular 3D resistive force theory (3D RFT) contact. Earth-trained policies keep 0.64 to 0.75 survival at the Moon on flat rigid ground in 10 s episodes, but falls continue, so longer episodes would see lower survival. Moon-trained and fine-tuned policies, in contrast, fall in every Earth episode, nearly all within 2 s. Rough-terrain training helps on uneven rigid ground only at training gravity. 3D RFT soil training helps on Earth sand but little on lunar soil. On three held-out soil models, Moon-trained policies lose no detectable progress at lunar gravity. Among tested strategies, only gravity-randomized rough 3D RFT training keeps mean survival above 0.83 at Earth and 0.95 at the Moon on all key tests up to 20 deg. Its ablation without soil training differs consistently only on soil at Earth gravity, falling in most sand episodes. These predictions depend on the contact model and soil stiffness, so lunar soil claims require agreement across soil models.","published":"2026-10-21","addedAt":"2026-10-21","year":2026,"venue":{"id":"preprint","label":"Preprint","year":null,"isWorkshop":false,"raw":null},"intersection":"robotics","domains":[],"techniques":[{"id":"surface-mobility","code":"T2","label":"Surface Mobility & Terramechanics","source":"official","via":["openalex-topic:Robotic Locomotion and Control","openalex-topic:Soil Mechanics and Vehicle Dynamics"]},{"id":"learning","code":"T8","label":"Learning-based Methods","source":"official","via":["openalex-topic:Reinforcement Learning in Robotics"]}],"missions":[],"indexedAs":{"arxiv":["cs.RO"],"openalexTopics":[{"id":"T10879","name":"Robotic Locomotion and Control","score":0.836},{"id":"T10462","name":"Reinforcement Learning in Robotics","score":0.156},{"id":"T11978","name":"Soil Mechanics and Vehicle Dynamics","score":0.003}],"ntrsSubjects":[],"ieeeTerms":[]},"cardTag":"RESEARCH // ORBITAL ROBOTICS","citations":0,"links":{"page":"https://space-robots.org/papers/paper/2610.30080","arxiv":"https://arxiv.org/abs/2610.30080","pdf":"https://arxiv.org/pdf/2610.30080","html":"https://arxiv.org/html/2610.30080","doi":null,"code":null,"project":null}}},{"rank":5,"note":"A single end-to-end RL policy for point-to-point quadruped jumps in lunar gravity, using only mission-compatible onboard sensing.","paper":{"id":"2610.29586","title":"End-to-End Reinforcement Learning for Quadrupedal Jumping in Lunar Gravity","authors":["Levin Müller","Oliver Fischer","Marco Trentini","Yusuke Tanaka","Marco Hutter"],"abstract":"Legged robots can extend planetary exploration to rugged and discontinuous terrain that is inaccessible to conventional rovers. In lunar gravity, jumping further expands this mobility by enabling robots to cross obstacles and gaps that cannot be traversed by walking alone. We present a multistage reinforcement-learning approach that produces a single end-to-end control policy for commanded point-to-point jumping under lunar gravity. The resulting policy has a simple architecture and relies exclusively on mission-compatible onboard sensing, making its deployment on a real lunar mission plausible. In simulation, the learned policy performs accurate point-to-point jumps with landing errors smaller than the robot's body size. We analyze the torque and power demands of lunar jumping and characterize performance degradation under extreme outlier events, including control-update freezes of up to \\SI{1}{s} during flight. Hardware transfer demonstrates that the policy can successfully execute a vertical jump under lunar-gravity offloading. Furthermore, the measured torque and power profiles are consistent with the simulation results and remain within the robot's hardware limits.","published":"2026-10-21","addedAt":"2026-10-21","year":2026,"venue":{"id":"preprint","label":"Preprint","year":null,"isWorkshop":false,"raw":null},"intersection":"both","domains":[{"id":"planetary-robotics","label":"PLANETARY ROBOTICS"}],"techniques":[{"id":"surface-mobility","code":"T2","label":"Surface Mobility & Terramechanics","source":"official","via":["openalex-topic:Robotic Locomotion and Control","openalex-keyword:quadrupedal robots","openalex-keyword:legged robots","title:quadruped"]},{"id":"learning","code":"T8","label":"Learning-based Methods","source":"official","via":["openalex-topic:Reinforcement Learning in Robotics","openalex-keyword:end-to-end reinforcement learning","title:reinforcement learning"]}],"missions":[],"indexedAs":{"arxiv":["cs.RO"],"openalexTopics":[{"id":"T10879","name":"Robotic Locomotion and Control","score":0.899},{"id":"T10462","name":"Reinforcement Learning in Robotics","score":0.099},{"id":"T11023","name":"Prosthetics and Rehabilitation Robotics","score":0}],"ntrsSubjects":[],"ieeeTerms":[]},"cardTag":"RESEARCH // PLANETARY ROBOTICS","citations":0,"links":{"page":"https://space-robots.org/papers/paper/2610.29586","arxiv":"https://arxiv.org/abs/2610.29586","pdf":"https://arxiv.org/pdf/2610.29586","html":"https://arxiv.org/html/2610.29586","doi":null,"code":null,"project":null}}},{"rank":6,"note":"OWLS plans rover paths directly over a VLM's dense feature map, turning a text prompt describing traversable terrain into a speed field for zero-shot navigation.","paper":{"id":"2610.32596","title":"OWLS: Open-world Language-conditioned Speed Fields for Zero-Shot Rover Navigation","authors":["Chahyon Ku","Erica Tevere","Georgios Georgakis","Rudra Mukherjee","and Bernadette Bucher"],"abstract":"Semantics can provide a proxy for physical traversability. Terrain that looks like loose sand or a rock field is likely to have similar physical properties. Open-vocabulary vision-language models (VLMs) enable this physical reasoning over terrain outside any fixed label set. We present Open-world Language-conditioned Speed Fields (OWLS), which plans directly over the dense feature map of a VLM. Prompted with a description of traversable terrain, the VLM scores every pixel of a single camera image; OWLS projects these scores onto the ground, adapts them to the platform with two morphological operations, and uses the result as the speed field of a minimum-time planner, so terrain appearance and obstacle clearance shape one trajectory. OWLS needs no task-specific training, no terrain annotation, no persistent map, and no exteroceptive sensor beyond the camera. We evaluate it on a physical rover indoors and off-road, on simulated Mars and Moon rovers over 200 procedurally generated terrains, and on real planetary imagery. Project page: https://chahyon-ku.github.io/owls/","published":"2026-10-23","addedAt":"2026-10-23","year":2026,"venue":{"id":"preprint","label":"Preprint","year":null,"isWorkshop":false,"raw":null},"intersection":"both","domains":[{"id":"planetary-robotics","label":"PLANETARY ROBOTICS"}],"techniques":[{"id":"planning-navigation","code":"T6","label":"Motion Planning & Navigation","source":"official","via":["openalex-topic:Robotic Path Planning Algorithms","openalex-keyword:rover navigation","title:rover navigation"]},{"id":"perception","code":"T5","label":"Perception & State Estimation","source":"official","via":["openalex-topic:Robotics and Sensor-Based Localization","openalex-keyword:open-vocabulary semantic segmentation"]},{"id":"foundation-models","code":"T9","label":"Foundation Models & LLMs","source":"official","via":["openalex-keyword:vision-language models","abstract:vision-language","abstract:vlms"]}],"missions":[],"indexedAs":{"arxiv":["cs.RO"],"openalexTopics":[{"id":"T10191","name":"Robotics and Sensor-Based Localization","score":0.316},{"id":"T10462","name":"Reinforcement Learning in Robotics","score":0.315},{"id":"T10586","name":"Robotic Path Planning Algorithms","score":0.283}],"ntrsSubjects":[],"ieeeTerms":[]},"cardTag":"RESEARCH // PLANETARY ROBOTICS","citations":0,"links":{"page":"https://space-robots.org/papers/paper/2610.32596","arxiv":"https://arxiv.org/abs/2610.32596","pdf":"https://arxiv.org/pdf/2610.32596","html":"https://arxiv.org/html/2610.32596","doi":null,"code":null,"project":null}}},{"rank":7,"note":"Head-to-head PPO vs. SAC for coordinated dual-arm space manipulator motion with inter-arm collision avoidance, motivated by debris removal and servicing.","paper":{"id":"2610.32034","title":"PPO vs. SAC: A Comparative Study for Cooperative Motion Control of Dual-Arm Space Robots","authors":["Nasim Fallahiarezoodar","Zheng H. Zhu"],"abstract":"The growing population of orbital debris has intensified interest in active debris removal and on-orbit servicing, for which dual-arm space manipulators offer improved grasping stability and load distribution relative to single-arm systems. However, coordinating two manipulators under inter-arm collision avoidance, orientation alignment, and sensing uncertainty remains challenging. This paper develops a coordinated reinforcement learning framework for cooperative motion planning of a dual-arm space manipulator in a target-reaching task with inter-arm collision-avoidance constraints. A centralized state representation combines joint-space, end-effector, and target-relative information, while a multi-objective reward function promotes target-reaching accuracy, orientation alignment, motion smoothness, synchronized task completion, and collision avoidance. Within this framework, Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC) are compared as representative on-policy and off-policy algorithms, and the learned policies are further evaluated for generalization to previously unseen target configurations. Across five random seeds, SAC substantially outperforms PPO within the considered fixed-base simulation scenarios: it reaches an 80% success threshold in every seed (median 1.6M environment steps) whereas PPO does not reach it within a 4M-step budget, and it attains a higher mean success rate (78.6% versus 7.4%) with lower position and orientation errors and tighter bimanual synchronization. Both algorithms degrade on unseen boundary targets, but SAC retains most of its performance while PPO's collapses. The results also expose a failure mode of PPO: under the two-phase shaped reward, it tends to converge to a reward-loitering local optimum rather than committing to the precision task. This behavior highlights the sensitivity of on-policy learning to reward shaping.","published":"2026-10-23","addedAt":"2026-10-23","year":2026,"venue":{"id":"aas-astrodynamics","label":"AAS/AIAA Astrodynamics 2026","year":2026,"isWorkshop":false,"raw":null},"intersection":"both","domains":[{"id":"in-space-servicing","label":"IN-SPACE SERVICING"},{"id":"orbital-robotics","label":"ORBITAL ROBOTICS"}],"techniques":[{"id":"learning","code":"T8","label":"Learning-based Methods","source":"official","via":["openalex-topic:Reinforcement Learning in Robotics","openalex-keyword:proximal policy optimization","title:ppo","abstract:reinforcement learning"]},{"id":"manipulation","code":"T1","label":"Manipulation & Grasping","source":"official","via":["openalex-topic:Robot Manipulation and Learning","abstract:space manipulator","abstract:grasp"]},{"id":"planning-navigation","code":"T6","label":"Motion Planning & Navigation","source":"official","via":["openalex-keyword:cooperative motion planning","abstract:collision avoidance","abstract:motion planning"]}],"missions":[],"indexedAs":{"arxiv":["cs.AI"],"openalexTopics":[{"id":"T10462","name":"Reinforcement Learning in Robotics","score":0.895},{"id":"T11701","name":"Space Satellite Systems and Control","score":0.1},{"id":"T10653","name":"Robot Manipulation and Learning","score":0.003}],"ntrsSubjects":[],"ieeeTerms":[]},"cardTag":"RESEARCH // IN-SPACE SERVICING","citations":0,"links":{"page":"https://space-robots.org/papers/paper/2610.32034","arxiv":"https://arxiv.org/abs/2610.32034","pdf":"https://arxiv.org/pdf/2610.32034","html":"https://arxiv.org/html/2610.32034","doi":null,"code":null,"project":null}}}]},"weeks":["2026-W44","2026-W43"]}