MEKAWIRE
Forschung

Studien

Neue Studien zu Robotik und Physical AI, jeweils mit ihrer Bedeutung für die Branche.

1 Studien
ThemaAlleManipulation 81Navigation 51Robotik allgemein 41VLA 33Roboterlernen 31Simulation 24Wahrnehmung 23Sicherheit 13Fortbewegung 12Humanoide 9Aerial 4World Action Models 4World-action model 3Aerial manipulation 2Sim-to-Real Manipulation 2Social Navigation 2Vision-Language Navigation 23D Assembly 13DGS dataset 15G Simulation 1A 100-hour wearable dataset of people walking and working with their hands, recorded as video, whole-body motion, finger motion and touch for humanoid training 1A 3D world action model pre-trained on 1.15 million human video episodes reaches 69.0% on ten simulated dexterous-hand tasks, 11.7 points above NVIDIA's GR00T N1.6 1A 4-billion-parameter perception model that points at and boxes objects precisely, used as the visual backbone for robot manipulation and driving policies 1A benchmark and 3.1k-demonstration dataset that tests whether humanoid robot policies can pick the right tool and then use it, standing still and while walking, on simulated and real Unitree G1 robots 1A controlled comparison of discrete-token and continuous action heads for single-task policies and PaliGemma-based VLA models, and why discrete heads fail or succeed 1A dual memory added to the pi0.5 VLA lifts real-robot success on four memory-dependent tasks from 37.5% to 62.5% and RoboMME benchmark success from 44.51% to 49.22% 1A frontier language model controlling a Franka arm learns from its own failed attempts by tracing late failures back to earlier decisions, with no demonstrations and no weight updates 1A general chat-style vision-language model drives a robot arm through mid-level commands and live feedback, with no robot training data 1A lightweight verifier that predicts the consequences of a VLA's planned actions and decides how many of them to execute before looking again 1A physics-based benchmark from KAIST that checks whether vision-language-action robot models crush or deform the objects they pick up, not only whether they finish the task 1A quadruped with no arm pushes two stacked boxes to a precise pose by crouching as it shoves; 90.1% success in simulation and zero-shot transfer to a Unitree Go1 1A reinforcement-learning reaching controller trained only in Isaac Lab that guides a UR10e arm to the stem of hydroponic lettuce without touching the growing channels 1A residual correction layer that keeps the size, energy and posture of humanoid gestures such as waving, bowing and crouching when an existing whole-body tracker executes them 1A safety layer that decides, at every control step, whether a flipping humanoid should finish the move, abort and land on its feet, or fall in a way that protects its head and hands 1A training add-on for VLA models such as OpenVLA-OFT and pi0.5 that keeps, rather than discards, background and lighting cues and lifts success under unseen conditions 1AMR Operations 1Active vision manipulation 1Adapting a frozen flow-matching VLA by learning only the noise it starts from 1Adaptive MPC 1Adaptive action horizons 1Adaptive intervention 1Adding a small latent world model to a pi0.5 policy so that predicted consequences of a planned action condition the action generator, without video generation or large-scale video pretraining 1Adding a tactile expert to the pi0.5 vision-language-action model so a robot learns contact-heavy tasks such as wiping, inserting and picking up chips from 50 demonstrations, without a separate tactile pretraining stage 1Aerial navigation 1Agricultural Robotics 1An object-centric VLA that tracks each object's identity and state and learns to predict how objects will move under the robot's actions 1Articulation Estimation 1Asking pi0.5 for an intent such as clearing a table instead of naming the action costs 15.2 points of task success, and people decompose multi-step requests less as tasks grow 1Audit of seven robot-learning benchmarks finds 22 bugs and 4 design flaws that make sped-up VLA models look better than the original 1Autonomous Driving 1Autonomous Exploration 1Battery-aware RL 1Belief-space planning 1Benchmark for adding wrist force sensing to VLA manipulation policies 1Benchmark for few-shot skill transfer with dexterous robot hands and tactile sensing 1Bimanual Manipulation 1Black-box correction of deployed robot policies with world-model planning 1Byzantine localization 1Canal inspection 1CapPro lets a pi0.5 robot policy attempt a new manipulation task after watching one human video, without new robot demonstrations or fine-tuning 1Co-speech Humanoid 1Coding agents that control robots and keep what they learn as reusable skills and lessons 1Compositional Manipulation 1Construction navigation 1Contact-Rich Manipulation 1Contact-rich Manipulation 1Continual reinforcement learning on top of robot foundation models 1Continuum sensing 1Controlled comparison of how to add world models to VLA robot policies 1Cross-embodiment transfer of humanoid whole-body controllers 1Cross-hand dexterous learning: a matched seven-hand, six-task simulation benchmark comparing action representations for one shared policy 1Curriculum RL 1Diagnosing why VLA robots fail at long household tasks that require moving around, and fixing planning, base positioning and grasping with agentic tools around an unchanged VLA 1Distributed Manipulation 1Distribution Shift 1Driving Simulation 1Efficient VIO 1Egocentric data 1Egocentric video 1Embodied Agents 1Embodied policy 1Empowerment MPC 1End-to-end drone navigation from stereo RGB images only, trained in simulation and flown on real hardware without fine-tuning 1Endoscopic navigation 1Endovascular robotics 1Evaluating robot policies without running them: fitted Q-evaluation on offline trajectories with a DINOv3 encoder and a vision-language success detector, against video world-model evaluators 1Evaluating robot world models against real executions 1Event-camera VO 1Exoskeletons 1Failure recovery 1Faster on-device inference for flow-matching vision-language-action models 1Federated 3D Perception 1Fine-tuning pretrained robot policies such as NVIDIA's GR00T N1.7 on a few demonstrations without losing robustness to new camera views, backgrounds and objects 1Fisheye detection 1Floorplan SLAM 1Force-aware robot foundation model pretrained on 2,000 hours of human tactile-glove data 1Geometry-aware manipulation 1Giving a VLA policy memory by feeding it compact latent codes of what changed between consecutive camera frames 1Grasping 1Humanoid control, generative modeling of policy weights 1Humanoid loco-manipulation 1Humanoid whole-body control 1Humanoid whole-body training data from robot-specific video generation 1Imitation Learning 1JEPA World Models 1Lab automation 1Laboratory Manipulation 1Language-to-motion for humanoids: event-grounded diffusion that generates Unitree G1 motion and edits it turn by turn 1Latent World Models 1Legged Manipulation 1Letting a humanoid revise its cart-pushing plan mid-motion as it learns the cart's mass and wheel constraints 1Letting a robot's vision-language-action model improve during deployment by reading its own recent attempts and rewards, without retraining any weights 1Letting humans and coding agents build and tune physics simulations of robot tasks from a phone video 1LiDAR Security 1Lifelong RL 1MAPF 1MAV guidance 1Making touch part of the robot's action: a shared code for motion and contact lifts real-world success on peeling, flipping, cutting and wiping by 21 to 36.5 points 1Manipulation LfD 1Marine simulation 1Memory-centric VLA 1Miniature Locomotion 1Mining Perception 1Mobile manipulation 1Model-based RL 1Monocular SSC 1Motion Planning 1Motion quality of VLA-driven arms: replaying identical commands and network delays through different low-level controllers to measure jerk against tracking error 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 1One reinforcement-learning policy that walks a 55 kg industrial quadruped over gaps, platforms, grating stairs and beams straight from raw LiDAR points, with no terrain map 1One reinforcement-learning policy that walks a Unitree Go2 with a 6-DoF arm and places the gripper to within 3.22 cm under simulated lunar gravity, with safety limits enforced during training 1One world model retrained from NVIDIA Cosmos 3 Edge that both predicts scenes and controls 20 robot families 1Open-weights VLA family for dual-arm robots: embodiment midtraining before task finetuning, plus online correction from a few human episodes 1Panoramic 3D Detection 1Panoramic segmentation 1Pedestrian tracking 1PersonNav 1Personalising a pretrained vision-language-action model to one home or office from a 20 to 30 minute phone scan, using a Gaussian-splat digital twin to generate pickup demonstrations automatically 1Planning 1Policy Optimization 1Post-training VLA policies on messy real data: a reachability graph scores which action chunks move a precise assembly task forward 1Predicting contact failures before they happen with a vision, action and force world model 1Preference Learning 1Pretraining a delivery robot's navigation model on 10,000 hours of first-person human walking video by predicting scene structure instead of pixels, then adapting it with 17 hours of robot driving 1ProWorld, a robot world model trained on 6,486 hours of data, predicts real-robot policy success with a 0.95 Pearson correlation, against 0.33 and 0.67 for Ctrl-World and Cosmos 3 Nano 1Quadruped locomotion 1Real-time VLA 1Real-to-sim 1Real-to-sim-to-real data generation for manipulation 1Real-world reinforcement learning that splits a dexterous robot hand's finger skill from the arm's coordination policy, so tool-use tasks such as working a power drill or a pipette can be learned in about an hour 1Reinforcement learning for industrial assembly 1Representation alignment 1Robot policies trained only on simulator-checked world-model videos reach 55.0% real-world success, 86% of what 50 human demonstrations deliver, and triple it in unfamiliar scenes 1Robot-Use Agents 1Runtime Verification 1Runtime recovery for a frozen VLA policy: a latent world model forecasts whether to keep going, move the camera or replan the arm 1Runtime safety for VLA policies: a training-free barrier filter that guards the gripper, wrist and forearm against a moving hazard on edge hardware 1Safe Generative Planning 1Safe RL-MPC 1Safety monitoring 1Safety-critical control 1Scissor-shaped handheld teaching tool that lets people feel grip forces while recording robot demonstrations, used to fine-tune pi0.5 on dual-arm Franka and Flexiv robots 1Shrinking a 3.3-billion-parameter robot foundation model into a 451-million-parameter student that runs in two inference steps 1Sim-to-real Navigation 1Sim-to-real manipulation 1Simulation geometry 1Skill Discovery 1Social navigation 1Soft Pneumatic Actuators 1Soft manipulation 1Sonar SLAM 1Splitting each two-handed robot action into separate left and right tokens that share a timestamp, so a pi0.5-based policy coordinates two dexterous arm-hand systems better 1Starting a robot policy's flow-matching generation from its own recent actions and frames instead of random noise, so vision-language-action and world-action models train in fewer epochs and act in one sampling step 1Streaming 3D reconstruction 1Submovement decomposition 1Surgical 1Surgical robotics 1Tactile Exploration 1Teaching a Unitree G1 humanoid martial-arts jumps and spins from video of national-level athletes, with fall recovery learned inside the same policy 1Teaching a real humanoid to return tennis balls using public walking data and ordinary tennis videos instead of new motion capture 1Teaching a soft-skinned humanoid face to move its eyes, neck and eyelids like a person: 27.9% lower closed-loop error than the best learned rival and about 80 times lower than scripted eye-head rules 1Temporal manipulation 1Terrain traversability 1Test-time verification for vision-language-action policies 1Testing whether failure alarms for robot AI models actually save time once false alarms and reset costs are counted 1Text-to-3D Policy 1Text-to-motion for humanoids: a body-part-aware latent diffusion model that writes Unitree G1 joint references a frozen tracker can execute 1Tool Co-Design 1Training a VLA for a drone with a robot arm entirely on synthetic trajectories, then chaining local skills across sites with a scene graph to complete multi-step missions 1Training a vision-language-action model to predict the physical effect of its next actions instead of the raw motor commands, then decoding actions from that effect 1Training care robots that help people stand up and transfer, using simulated humans who move back 1Training household mobile manipulators with targeted human corrections instead of more demonstrations 1Trajectory Correction 1Transferring task structure from human videos to robot world-action models 1Traversability 1Turning a safety shield's would-be corrections into training data: a pi0.5 policy gains 8.5 points of success and causes 35.6% less contact cost, with no shield at deployment 1Turning first-person human demonstrations into training data a humanoid can use directly, so a vision-language-action model trained without any robot demonstrations can carry objects about 10 m on a Unitree G1 1Turning ordinary human videos into skill-labelled training data for a dexterous robot hand 1Turning phone-recorded human demonstrations into per-robot post-training data by retargeting object motion 1Turning recorded human hand-object demonstrations into physically valid robot-hand trajectories with one shared reinforcement-learning policy instead of one optimisation per clip 1Turning third-person human videos of table-tennis serves, pickups and ladder climbing into Unitree G1 skills by preserving where and how fast contacts happen 1Two Unitree G1 humanoids playing table-tennis doubles with learned collision-risk filtering 1UAV Navigation 1UAV Swarms 1UAV traffic monitoring 1Underwater Robotics 1Underwater active mapping 1Underwater perception 1Unified world-action model 1Using a learned multi-view world model as the training environment for reinforcement learning on a pi0.5 policy, so the robot improves with little extra real-world interaction 1Using a video generator to turn four real clips of a person carrying a box into 256 synthetic interaction videos, then into simulation training data for a humanoid that picks up, carries and drops unseen objects 1VISTA, a touch-and-vision robot policy, reaches 80.9% success on eight contact-rich simulated tasks from 50 demonstrations each and 22 of 30 real trials against 4 of 30 for the strongest baseline 1VLA Conditioning 1VLA Defense 1VLA Robustness 1VLA inference efficiency 1VLA manipulation 1VLA robustness 1VLA robustness benchmark 1Video world models 1Vision-language-action models without wrist cameras (learned foveation) 1Visual Generalization 1Visual Grounding 1Visual Reward Models 1Visuomotor Fusion 1Whether labelling messy human demonstrations by quality improves robot foundation models 1Whole-body mobile manipulation and teleoperation 1Why humanoid reinforcement-learning pipelines report success that is not there: failures hidden in curriculum gates, evaluation scripts and reference motions as well as in the reward, with measured examples from one simulated humanoid 1World Models 1World models 1World-Action Models 1World-action model that retrains on its own successful, mediocre and failed real-robot rollouts and improves round after round 1compliant humanoid whole-body control 1data engine converting egocentric human video into replayable bimanual dexterous robot demonstrations 1dexterous in-hand manipulation with learned world models 1efficient inference for vision-language-action robot policies 1humanoid learning from human video 1humanoid teleoperation with heavy unknown payloads 1humanoid whole-body manipulation 1large-scale tactile demonstration dataset and robot-free data collection 1open vision-language-action model with register-token memory 1post-training vision-language-action policies for industrial disassembly 1real-world reinforcement learning for robot policies 1robot foundation model training 1robot learning from human video 1tactile pretraining for vision-language-action models 1terrain-aware behavioral foundation model for humanoid whole-body control 1training-signal allocation for vision-language-action robot policies 1vision-language-action models 1world models for dexterous manipulation 1
NiveauAlleEinstiegFortgeschrittenExperten
Gefiltert nach Technologie Imitationslernen, zurücksetzen
arXiv
Turning ordinary human videos into skill-labelled training data for a dexterous robot handFortgeschritten

DexStitch: Geschickte Handfertigkeiten aus Menschenvideos lernen und verketten, mit kontaktbewusstem Robot-Inpainting

Byeonggyeol Choi, Seongmin Jung, Unghui Nam, Youngbo Song, Sunhong Min, MinHyung Lee, Byonghyo Shim, Yongseok LeeSeoul National University, Hyundai Motor Company Robotics Lab, DGIST

Ein Preprint der Seoul National University, des Robotics Lab von Hyundai Motor und des DGIST erreicht mit einer Roboterhand 77,5 Prozent Erfolg bei Abfolgen aus vier Haushaltsfertigkeiten, während die mit denselben Daten trainierte Vergleichsmethode UniSkill auf 0 Prozent kommt. Der Kniff: Die Bewegung zwischen zwei Fertigkeiten wird als eigene Fertigkeit gelernt, aus rund 240 Menschenvideos und nur 32 kurzen Teleoperationsläufen.

FeinmanipulationImitationslernenTeleoperation
Hyundai Motor GroupRainbow RoboticsROBOTERA
Vollständige Studie lesen (PDF)

Themenseiten

Manipulation 6World Action Models 5World-action model 3Sim-to-Real Manipulation 3Social Navigation 3

Die Studien stammen aus den Robotik-Feeds von arXiv. Solange unsere Zusammenfassung fehlt, erscheinen die ersten Zeilen des Abstracts.