MEKAWIRE
Research

Papers

New robotics and physical-AI papers, with what each one means for the industry.

415 papers
TopicAllManipulation · 76VLA · 66Navigation · 35Robotics · 32Learning · 27Perception · 22Humanoid · 21Simulation · 15Locomotion · 14Safety · 11World Action Models · 4World-action model · 3Aerial · 2Aerial manipulation · 2Dexterous manipulation · 2Sim-to-Real Manipulation · 2Social Navigation · 2Vision-Language Navigation · 2World models · 23D Assembly · 13DGS dataset · 15G Simulation · 1AMR Operations · 1Active vision manipulation · 1Adaptive MPC · 1Adaptive action horizons · 1Adaptive intervention · 1Aerial Manipulation · 1Aerial navigation · 1Agricultural Robotics · 1Articulation Estimation · 1Autonomous Driving · 1Autonomous Exploration · 1Battery-aware RL · 1Belief-space planning · 1Bimanual Manipulation · 1Byzantine localization · 1Canal inspection · 1Co-speech Humanoid · 1Compositional Manipulation · 1Construction navigation · 1Contact-Rich Manipulation · 1Contact-rich Manipulation · 1Continual reinforcement learning on top of robot foundation models · 1Continuum sensing · 1Cross-embodiment transfer of humanoid whole-body controllers · 1Curriculum RL · 1Dexterous Manipulation · 1Distributed Manipulation · 1Distribution Shift · 1Driving Simulation · 1Efficient VIO · 1Egocentric data · 1Egocentric video · 1Embodied Agents · 1Embodied policy · 1Empowerment MPC · 1Endoscopic navigation · 1Endovascular robotics · 1Evaluating robot world models against real executions · 1Event-camera VO · 1Exoskeletons · 1Failure recovery · 1Federated 3D Perception · 1Fisheye detection · 1Floorplan SLAM · 1Geometry-aware manipulation · 1Grasping · 1Grounding · 1Humanoid Control · 1Humanoid RL · 1Humanoid Tool Use · 1Humanoid control · 1Humanoid loco-manipulation · 1Humanoid safety · 1Humanoid whole-body control · 1Imitation Learning · 1JEPA World Models · 1Lab automation · 1Laboratory Manipulation · 1Latent World Models · 1Legged Manipulation · 1LiDAR Security · 1Lifelong RL · 1MAPF · 1MAV guidance · 1Manipulation LfD · 1Manipulation Safety · 1Marine simulation · 1Memory-centric VLA · 1Miniature Locomotion · 1Mining Perception · 1Mobile manipulation · 1Model-based RL · 1Monocular SSC · 1Motion Planning · 1Multi-Agent RL · 1Multi-Robot Exploration · 1Multi-Robot Manipulation · 1Multi-UAV Exploration · 1Multi-UAV State Estimation · 1Multi-agent RL · 1Multi-agent motion planning · 1Multi-camera odometry · 1Multi-fidelity RL · 1Multi-robot manipulation · 1Multi-skill Manipulation · 1Multi-view perception · 1Multimodal simulation · 1Navigation & Mapping · 1Object Pose Estimation · 1Object Search · 1Panoramic 3D Detection · 1Panoramic segmentation · 1Pedestrian tracking · 1PersonNav · 1Planning · 1Policy · 1Policy Optimization · 1Preference Learning · 1Quadruped locomotion · 1Real-time VLA · 1Real-to-sim · 1Reinforcement learning for industrial assembly · 1Representation alignment · 1Robot-Use Agents · 1Runtime Verification · 1Safe Generative Planning · 1Safe RL-MPC · 1Safe VLA · 1Safety monitoring · 1Safety-critical control · 1Sim-to-real Navigation · 1Sim-to-real manipulation · 1Simulation geometry · 1Skill Discovery · 1Social navigation · 1Soft Pneumatic Actuators · 1Soft manipulation · 1Sonar SLAM · 1Streaming 3D reconstruction · 1Submovement decomposition · 1Surgical · 1Surgical robotics · 1Tactile · 1Tactile Exploration · 1Tactile VLA · 1Temporal manipulation · 1Terrain traversability · 1Text-to-3D Policy · 1Tool Co-Design · 1Trajectory Correction · 1Traversability · 1UAV Navigation · 1UAV Swarms · 1UAV traffic monitoring · 1Underwater Robotics · 1Underwater active mapping · 1Underwater perception · 1Unified world-action model · 1VLA Conditioning · 1VLA Defense · 1VLA Robustness · 1VLA evaluation · 1VLA manipulation · 1VLA robustness · 1Video world models · 1Visual Generalization · 1Visual Grounding · 1Visual Reward Models · 1Visuomotor Fusion · 1Visuotactile manipulation · 1World Models · 1World-Action Models · 1
LevelAnyBeginnerIntermediateExpert
arXiv
ManipulationExpert

Keep the Effect, Drop the Actor: Programmable Effect-to-Execution World-Action Models

Junyi Hu, Zhewen He, Zhenhua Li, Yi Fang

PEWAM is a preprint describing a 71.5M-parameter world-action model that compiles robot demonstrations into editable “effect programs”—3D object keypoint trajectories, grasp contact points, and final scene configuration—rather than actor-specific motions. On held-out LIBERO-Goal tasks, one program from a single demonstration completes 40 of 90 episodes, while the same backbone under goal-image or language conditioning and prior demonstration-conditioned methods complete at most 19. This matters because turning demonstrations into coordinate-based programs enables editing and cross-robot-arm reuse, a potential route to more flexible robot programming.

Read the full paper (PDF)
arXiv
Geometry-aware manipulationExpert

Robot Learning on Discrete Surfaces: Theory and Applications

Matteo Dalle Vedove, Fares J. Abu-Dakka, Luigi Palopoli, Daniele Fontanelli, Matteo Saveriano

Preprint. The authors build a discrete Riemannian framework that lets robots learn motion and functions directly on polyhedral surface meshes rather than treating surfaces as external constraints. They apply it to Dynamic Movement Primitives, Gaussian Processes, and Riemannian Flow Matching, showing stability improvements over MeshDMP and roughly halving RFM training time. This matters for any contact-rich surface task such as polishing, drawing, or inspection on CAD or reconstructed meshes.

Read the full paper (PDF)
arXiv
Compositional ManipulationExpert

Rethinking World-Action Model for Compositional and In-Context Robotic Manipulation

Shukai Gong, Xuanran Zhai, Yintianrun Zhang, Ruopeng Cui, Ye Huang, Yiyang Fu, Dexuan Lyu, Chaojie Li, Xinyi Song, Peiwen Lin, Chuang Wang, Mingyuan Jia, Yufan Deng, Jiaxin Fang, Bo Liang, Jiaxin Li, Yuxiang Gao, Hao Liu, Daquan Zhou

This preprint introduces ViGAR, a hierarchical manipulation model that separates visual subgoal planning from subgoal-conditioned world-action generation, using a shared pretrained Cosmos3-Nano world model. On the RoboTwin Clean2Random benchmark it reaches 82.00% Clean and 67.02% Random success, 12.86 points above the best baseline average. It matters because image-based subtask goals allow the same policy to be steered to new task compositions via a global goal image without retraining.

Read the full paper (PDF)
arXiv
Policy OptimizationExpert

FERPO: Forward Entropy-Regularized Policy Optimization

Sebastian Sanokowski, Alireza Sarmadi, Majid Khadiv

This arXiv preprint introduces FERPO, an on-policy maximum-entropy reinforcement-learning algorithm that updates a policy from critic values without taking critic action derivatives. It roughly matches REPPO on DMC and ManiSkill while reaching high scores with fewer environment interactions, and at K=16 its uncached actor-critic updates are 6.9% faster than REPPO, or 22.2% faster with cached actor updates. The method matters for massively parallel robot-policy training because it avoids unreliable learned action gradients and improves wall-clock and sample efficiency.

RL
Read the full paper (PDF)
arXiv
Humanoid loco-manipulationExpert

InterEvolve: Test-Time Evolution of Reward Programs for Humanoid Loco-Manipulation

Zhuo Lin, Sirui Xu, Liuyu Bian, Yu-Xiong Wang, Liang-Yan Gui

InterEvolve is a preprint framework that evolves staged reward programs at test time so a frozen humanoid controller can solve new loco-manipulation tasks without retraining. It raises average success from 34.6% for the best calibrated fixed reward to 86.5% across eight goal-conditioned task families and transfers evolved skills to a real Unitree G1 using onboard perception. This turns task adaptation into a search over objectives rather than policy retraining, with more test-time compute improving results.

Humanoid
Read the full paper (PDF)
arXiv
Multi-agent motion planningExpert

Training-Free Diffusion Planning with Analytical Local Scores

Michael Y. Fatemi, Jinhao Liang, Ferdinando Fioretto

This arXiv preprint introduces TFDP, a training-free 2D multi-agent path planner that replaces learned diffusion scores with analytical local scores for smoothness, obstacle avoidance, velocity, and inter-agent separation. On standard benchmarks it reaches 96–100% success while averaging 0.59 s per plan versus 88.23 s for the strongest diffusion baseline, and it scales to 300 agents and 100 obstacles in 5.39 s on an A100 GPU. The work matters because it retains diffusion-style iterative refinement without needing trajectory datasets or map-specific training.

Read the full paper (PDF)
arXiv
Multi-fidelity RLExpert

Multi-Fidelity Policy Gradients Stabilize Data-Scarce Reinforcement Learning

Xinjie Liu, Ruihan Zhao, Anirban Chaudhuri, Cyrus Neary, Ufuk Topcu, David Fridovich-Keil

This preprint introduces MFPG-PPO, an algorithm that uses abundant low-fidelity simulator data as a control variate to reduce variance in scarce high-fidelity policy-gradient updates. On the hardest simulated locomotion task it matches PPO trained with 16 times more high-fidelity data, and on a physical Franka arm it reaches 91.7% success with only 4 real episodes per update, versus 55.0% for PPO on high-fidelity data alone. If validated, the approach could make on-policy robot learning much more sample-efficient in real-world deployments.

RL
Read the full paper (PDF)
arXiv
Multimodal simulationExpert

TouchTherm: Building Multimodal Digital Twins of Objects for Tactile and Thermal Rendering

Yitao Zhang, Hong Ying, Haoran Guo, Xiaoying Zhou, Guanyu Chen, Chenxi Xiao

TouchTherm is a preprint framework that turns real objects into simulation-ready digital twins combining a coarse visual mesh, registered tactile micro-height fields, and a physics-regularized dynamic thermal field. On 20 objects it predicts held-out surface temperature with MAE 0.465°C at 30 s and 0.592°C at 45 s, and improves synthetic-to-real tactile object recognition from 20.0% to 34.0% Top-1 accuracy over coarse geometry. The work matters because it closes part of the simulation-to-real gap for contact-rich robotics and VR by making tactile and thermal cues queryable in simulation.

Read the full paper (PDF)
arXiv
Embodied AgentsExpert

Reconstruct, Practice, Go Real: Guided Self-Improvement for Embodied Agents

Yen-Jen Wang, Haozhe Jiang, Shuying Deng, Haoru Xue, Weirui Ye, Rocky Duan, Nika Haghtalab, S. Shankar Sastry, Pieter Abbeel, Haozhi Qi

This preprint introduces RPG, a framework that uses offline demonstration videos, simulator practice, and privileged-state rollouts to autonomously improve reusable robot skills and prompts without updating model weights. On held-out initializations of 22 manipulation tasks, RPG reaches 95.0% success after 15 practice rounds, beating ASPIRE at 75.5% and CaP-Agent0 with GPT-6 Astra Pro at 60.0%, and succeeds in all 30 physical trials across three tasks. The result matters because it shows shared skill code and a system prompt can be a practical target for autonomous robot capability improvement.

Read the full paper (PDF)
arXiv
Aerial manipulationExpert

AFD-CAMLs: Agile Force-Distribution-Aware Planning and Control for Cable-Suspended Aerial Multi-Lifting Systems

Antreas Kourris, Sihao Sun

This preprint presents AFD-CAMLs, a hybrid planning and control framework for cable-suspended payload transport with four or more UAVs. It combines null-space force-allocation planning, force-aware NMPC, and an admittance filter that uses onboard tension estimates to balance cable loads. In agile real-robot flights, it improved 3D payload tracking error from 0.179 m to 0.153 m and reduced aggregate tension-tracking RMSE from 3.51 N to 2.09 N versus the baseline, without sacrificing agility.

Read the full paper (PDF)
arXiv
Active vision manipulationExpert

ActiveWAM: Evidence-Aware Active Vision for World-Action Models

Renjun Wu, Luzhou Ge, Xuesong Li

ActiveWAM is a preprint describing a world-action model that jointly controls a dual-arm robot and a two-degree-of-freedom pan/tilt camera, learning when to keep a useful view and when to move to gather new evidence. It reaches 80.3% clean and 53.3% compound-shift success on the 50-task RoboTwin-AV benchmark, versus 76.3% and 33.3% for Fast-WAM, and completes 50.0% of two-stage physical kitchen tasks versus 23.3% for Fast-WAM. This matters because it removes test-time inversion and viewpoint annotations while improving active-vision manipulation under distribution shift.

Read the full paper (PDF)
arXiv
Underwater perceptionExpert

SonarVoxNet: Diver Detection in 3D Bounding Box using 3D Sonar

Eugene Park, Jiwon Lee, Seyoung Kan, Trung Dong, Xiaomin Lin, Jane Shin

This preprint introduces SonarVoxNet, a 3D sonar detector that predicts full 9-degree-of-freedom oriented bounding boxes for divers, plus Diver3D, a public 37,010-frame dataset with full 3D orientation labels. With augmentation and causal refinement, SonarVoxNet reaches 0.774 [email protected] and 0.719 SDS on the held-out test set, far above LiDAR-derived baselines that top out at 0.374 [email protected]. The work shows that pitch and roll of a freely swimming diver can be recovered from sparse acoustic returns, which matters for AUV diver-robot interaction.

LiDAR
Read the full paper (PDF)
arXiv
Belief-space planningExpert

BLT*: Informed Belief Localization Trees for Uncertainty-Aware Planning on Digital Twins

Elliot Preston-Krebs, Abhishek Goudar, Timothy D. Barfoot

Preprint. The authors introduce Informed BLT*, a sampling-based belief-space planner that extends RRT*/Informed RRT* using the 2-Wasserstein metric for isotropic Gaussian beliefs. Across 16 simulated and digital-twin maps, BLT* found first solutions faster than B-RRT, B-SST and RRBT in nearly all maps—for example, 0.09 s versus 2.38–5.71 s on a 205×124 m campus digital twin—while solving a larger office twin where baselines mostly failed. This matters for ground robots that must plan paths that stay well-localized, not just collision-free.

Read the full paper (PDF)
arXiv
5G SimulationExpert

Network-in-the-Loop at Scale: GPU-Batched 5G Simulation for Massively Parallel Robot Learning

Zifan Zhang, Mingzhe Han, Kannan Athreya, Yuchen Liu

Isaac-Net is a preprint describing a GPU-batched 5G New Radio network simulator that runs in lockstep with Isaac Lab physics, enabling thousands of parallel robot-learning environments to share realistic uplink contention. It reproduces ns-3 5G-LENA median delay within roughly 5 to 10 percent, low on an unseen carrier and high at 32 robots per environment in closed loop, and captures Age of Information better than an independent per-message delay. This matters for training multi-robot policies that need to learn network-aware coordination on private 5G fleets.

Read the full paper (PDF)
arXiv
Visual GroundingExpert

GroundAnything: Reconciling Parallel Decoding with Precise Visual Grounding at Flash Speed

Qize Yu, Lianrui Fan, Bowen Ping, Xini Ding, Zetian Song, Junbo Niu, Kaixuan Wang, Tianxing Chen, Yue Chen, Minghua He, Yuran Wang, Jie Huang, Haojun Zhang, Min Chen, Hao Li, Wenxuan Song, Ruihai Wu, Xianming Liu, Shilong Liu, Shuchang Zhou

Preprint. The authors introduce GroundAnything, a 4B-parameter visual grounding model that uses bidirectional diffusion with blockwise denoising for parallel spatial decoding instead of sequential autoregressive token generation. Its autoregressive variant, GroundAnything-VLM, reports 72.42% across 30 grounding benchmarks, ahead of GPT-6 Astra at 71.35%. This matters for latency-sensitive robotics and interactive vision systems that need fast, precise localization.

Read the full paper (PDF)
arXiv
Tool Co-DesignExpert

Tool-Policy Co-Design for Powder Weighing in Laboratory Automation

Nikola Radulov, Xin Yang, Kevin S. Luck, Gabriella Pizzuto

This preprint presents a co-design method that jointly optimizes a powder-dispensing spoon's geometry and its reinforcement-learning control policy for robotic laboratory weighing. In real-robot trials, the best co-designed tool reduced overall weighing error to 2.23±3.00 mg from 4.09±6.82 mg for a standard tool across seven powders, about a 45% relative reduction. The approach addresses a bottleneck in self-driving labs by adapting tools, not just policies, to material flow behavior.

Read the full paper (PDF)
arXiv
LiDAR SecurityExpert

TACTIC: Temporal and Context-Aware LLM Tactical Planning for Roadside LiDAR Attacks

Yiming Gao, Shaocheng Luo

This preprint presents TACTIC, a scene-aware framework that uses a multimodal large language model and roadside perception to choose and tune physical LiDAR attacks in dynamic traffic. In 280 randomized CARLA trials, the full system caused collisions in 100% of evaluation cases, versus 35% for a fixed rule and 60% for random attack selection. The result shows attack success depends on scene context, which matters for evaluating LiDAR safety failures in autonomous driving.

LiDAR
Read the full paper (PDF)
arXiv
ManipulationExpert

DynaHarness: A Dynamic Physical Harness for Self-Evolving Robot Agents

Haoyuan Deng, Jiebin Liu, Tengxiao Zhang, Langning Yan, Hongye Cao, Ziwei Wang

DynaHarness is an arXiv preprint that adds a dynamic physical harness around a frozen vision-language-action policy: a slow Qwen3-VL-4B planner proposes capabilities, while a fast 2 Hz governor grounds, monitors, substitutes and recovers them. On LIBERO-Pro it reaches 75.2% success on 800 newly sampled states versus 17.5% for the frozen π0.5 policy, and its self-evolution gains transfer to new states. This gives robot-agent builders a retraining-free way to govern and improve existing policies from execution evidence.

VLA
Read the full paper (PDF)
arXiv
Sonar SLAMExpert

BatSLAM 2.0: Sequence-Verified Sonar Place Recognition in a Robust Pose Graph

Jan Steckel

BatSLAM 2.0 is a preprint describing a sonar-only simultaneous localization and mapping system that uses a biomimetic binaural acoustic front-end, sequence-verified place recognition, and a factor-graph pose graph. In 12 simulated long-range drives with calibrated odometry, it produced consistent maps with no wrong loop closures and a post-alignment trajectory error of 0.18–0.53 m; on a real beamformed sonar recording, it cut error from 3.62 m to 0.67 m. This matters for robots that must operate in darkness, smoke, or dust where cameras and lidar fail.

Read the full paper (PDF)
arXiv
Safe VLAExpert

Learning from Runtime Feedback through Failure-Bank Self-Evolution for Vision-Language-Action Models

Mingyue Cui, Zheyuan Liu, Yihan Zhu, Zheyuan Zhang, Meng Jiang

This preprint presents FailBank, a framework that turns an observe-only safety teacher's runtime corrections into training records for vision-language-action policies. On VLA-Arena static-obstacle tasks, it raises task success by 8.5 and 6.9 percentage points over base policies for two VLA backbones while cutting policy-induced cumulative cost by 35.6% and 23.8%. The method matters because it lets robot policies learn a persistent balance between completing a task and avoiding unintended contact, instead of relying on temporary runtime shields.

VLA
Read the full paper (PDF)

Topic pages

Manipulation · 76VLA · 65Navigation · 35General robotics · 32Robot learning · 27Perception · 22Humanoid · 21Simulation · 15Locomotion · 13Safety · 11World Action Models · 5World-action model · 3Aerial manipulation · 3Dexterous manipulation · 3Sim-to-Real Manipulation · 3Social Navigation · 3

Papers come from arXiv robotics feeds. Where our summary is not written yet, you see the opening lines of the abstract.