robot-learning

Hugging Face launched LeLab, a graphical user interface for LeRobot that enables training AI robots without command-line interaction, featuring zero-terminal setup, data collection, and one-click GPU training via Hugging Face Jobs.

GRAIL generates diverse humanoid manipulation and locomotion data using 3D assets and video foundation models, enabling effective sim-to-real transfer for humanoid robot control with high real-world success rates.

This paper introduces Parameterized Diffusion Policy (PDP), a framework that makes diffusion policies controllable by conditioning on low-dimensional latent parameters, enabling smooth behavior interpolation and adaptation without retraining. It demonstrates improved performance on complex multimodal robot tasks in simulation and real-world experiments.

WALL-WM advances video-action learning by using semantic events as learning units instead of fixed action chunks, enabling more flexible and scalable vision-language-action training and inference.

Proposes the Agentic-VLA framework, introducing agents into the VLA loop, enabling the vision-language-action model to self-evolve and surpass existing leading embodied models on all metrics.

This paper proposes a thinking-learning interaction model for autonomous robots, enabling them to adaptively discover new features, expand output categories, update learning models, and reconstruct action routines in open environments. Experimental results demonstrate significant improvements in recognition accuracy, category formation, and action efficiency.

Hugging Face released LeRobotHF, an open-source humanoid robot platform with hardware, simulation, and training tools, costing roughly $2,500 to build.

Hugging Face released LeRobot Humanoid, an open-source, low-cost ($2,500) 3D-printed humanoid robot platform, including hardware, simulation, and control tools for robot learning research.

Hugging Face has released LeRobot Humanoid, an open-source, low-cost ($2.5k) 3D-printed humanoid robot designed for robot learning, including hardware, software, simulation, and training tools.

LeRobot Humanoid is an open, low-cost bipedal robot platform for robot learning, including hardware, runtime, identification tools, and training environments, built from 3D-printed and off-the-shelf parts for about $2,500.

Boston Dynamics demonstrates how the new Atlas humanoid robot learns complex manipulation tasks like lifting heavy refrigerators through simulation training, enabling rapid iteration from design to real-world execution with minimal simulation-to-reality gap.

DexJoCo introduces a benchmark and toolkit for task-oriented dexterous manipulation in MuJoCo, featuring 11 functional tasks, a low-cost data collection system, and comprehensive evaluations that highlight limitations in current dexterous manipulation policies.

PhysBrain 1.0 is a technical report presenting a method that uses human egocentric video to generate physical commonsense supervision for vision-language-action models, achieving state-of-the-art results on embodied control benchmarks including ERQA, PhysBench, SimplerEnv-WidowX, LIBERO, and RoboCasa.

FrameSkip is a data-layer frame selection method that improves Vision-Language-Action (VLA) policy training by prioritizing high-importance frames based on action variation and visual-coherence metrics, achieving a macro-average success rate of 76.15% across three benchmarks while using only 20% of unique frames.

This paper introduces Domain-Randomized Instance Set (DRIS), a method that simultaneously represents multiple randomized instances to improve sim-to-real transfer for dexterous manipulation. It demonstrates zero-shot transfer on a challenging reactive catching task with a flat plate end-effector, requiring no real-world fine-tuning.

This paper introduces RLA-WM, a visual feature-based world model that leverages residual latent actions and flow matching to efficiently predict future visual states. The method outperforms existing video-diffusion and feature-based approaches while enabling novel robot learning techniques from offline, actionless demonstration videos.

This paper introduces TT4D, a novel pipeline and large-scale dataset for reconstructing table tennis gameplay in 4D from monocular videos. It features a unique lift-first approach that estimates 3D ball trajectories and spin before time segmentation, enabling robust reconstruction even with occlusions.

This comprehensive survey reviews the literature on world models for robot learning, covering their roles in policy learning, planning, and simulation. It highlights key paradigms, benchmarks, and future directions for predictive modeling in embodied agents.

NVIDIA highlights breakthroughs in physical AI and robotics during National Robotics Week, announcing new technologies including NVIDIA Isaac GR00T open models for natural language instruction understanding, Cosmos world models for synthetic data generation, Newton 1.0 physics engine, and expanded simulation capabilities with Isaac Sim 6.0 and Isaac Lab 3.0 to accelerate robot development from training to real-world deployment.

The article discusses the challenges of programming robots due to Moravec's paradox and proposes robot learning as a solution to enable embodied intelligence.