In-Hand Object Rotation via Rapid Motor Adaptation

2022/10/10 by Haozhi Qi, Ashish Kumar, Qi, Haozhi +7 · 41 citations
Computer Science · Engineering · #Reinforcement Learning in Robotics #Robot Manipulation and Learning #Robotic Locomotion and Control #cs.AI #cs.CV #cs.LG #cs.RO

paper · pdf · doi:10.48550/arxiv.2210.04887

CoRL 2022. Code and Website: https://haozhi.io/hora

arxiv created 2022/10/10 · arxiv updated 2022/10/11

Abstract

Generalized in-hand manipulation has long been an unsolved challenge of robotics. As a small step towards this grand goal, we demonstrate how to design and learn a simple adaptive controller to achieve in-hand object rotation using only fingertips. The controller is trained entirely in simulation on only cylindrical objects, which then - without any fine-tuning - can be directly deployed to a real robot hand to rotate dozens of objects with diverse sizes, shapes, and weights over the z-axis. This is achieved via rapid online adaptation of the controller to the object properties using only proprioception history. Furthermore, natural and stable finger gaits automatically emerge from training the control policy via reinforcement learning. Code and more videos are available at https://haozhi.io/hora

Cited by

• Concurrent Prehensile and Nonprehensile Manipulation: A Practical Approach to Multi-Stage Dexterous Tasks
• World Models for Learning Dexterous Hand-Object Interactions from Human Videos
• Learning Dexterous Manipulation Skills from Imperfect Simulations
• Beyond Topology: A Morphological Symmetry Graph Representation for Locomotion Policy Learning
• METIS: Multi-Source Egocentric Training for Integrated Dexterous Vision-Language-Action Model
• Enhancing Tactile-based Reinforcement Learning for Robotic Control
• Implicit State Estimation via Video Replanning
• Phys2Real: Fusing VLM Priors with Interactive Online Adaptation for Uncertainty-Aware Sim-to-Real Manipulation
• DexNDM: Closing the Reality Gap for Dexterous In-Hand Rotation via Joint-Wise Neural Dynamics Model
• AVO: Amortized Value Optimization for Contact Mode Switching in Multi-Finger Manipulation
• Diffusing Trajectory Optimization Problems for Recovery During Multi-Finger Manipulation
• AD-NODE: Adaptive Dynamics Learning with Neural ODEs for Mobile Robots Control
• In-Hand Manipulation of Articulated Tools with Dexterous Robot Hands with Sim-to-Real Transfer
• Learning Geometry-Aware Nonprehensile Pushing and Pulling with Dexterous Hands
• LodeStar: Long-horizon Dexterity via Synthetic Data Augmentation from Human Demonstrations
• Text2Touch: Tactile In-Hand Manipulation with LLM-Designed Reward Functions
• Solving Robotics Tasks with Prior Demonstration via Exploration-Efficient Deep Reinforcement Learning
• HERMES: Human-to-Robot Embodied Learning from Multi-Source Motion Data for Mobile Dexterous Manipulation
• Robot Trains Robot: Automatic Real-World Policy Adaptation and Learning for Humanoids
• DexReMoE:In-hand Reorientation of General Object via Mixtures of Experts
• Learning Pivoting Manipulation with Force and Vision Feedback Using Optimization-based Demonstrations
• A roadmap for AI in robotics
• SimLauncher: Launching Sample-Efficient Real-world Robotic Reinforcement Learning via Simulation Pre-training
• Scaffolding Dexterous Manipulation with Vision-Language Models
• The MOTIF Hand: A Robotic Hand for Multimodal Observations with Thermal, Inertial, and Force Sensors
• ViTacFormer: Learning Cross-Modal Representation for Visuo-Tactile Dexterous Manipulation
• Tactile Beyond Pixels: Multisensory Touch Representations for Robot Manipulation
• Touch begins where vision ends: Generalizable policies for contact-rich manipulation
• CARoL: Context-aware Adaptation for Robot Learning
• Morphologically Symmetric Reinforcement Learning for Ambidextrous Bimanual Manipulation
• DexMachina: Functional Retargeting for Bimanual Dexterous Manipulation
• DexUMI: Using Human Hand as the Universal Manipulation Interface for Dexterous Manipulation
• Situationally-Aware Dynamics Learning
• Guided Policy Optimization under Partial Observability
• Composing Dextrous Grasping and In-hand Manipulation via Scoring with a Reinforcement Learning Critic
• Rapid Embodiment Adaptation for Quadrupedal Locomotion
• Distilling Realizable Students from Unrealizable Teachers
• DexCtrl: Towards Sim-to-Real Dexterity with Adaptive Controller Learning
• DexDrummer: In-Hand, Contact-Rich, and Long-Horizon Dexterous Robot Drumming
• Towards Human-level Dexterous Teleoperation
• RoboVerse: Towards a Unified Platform, Dataset and Benchmark for Scalable and Generalizable Robot Learning
• Crossing the Human-Robot Embodiment Gap with Sim-to-Real RL using One Human Demonstration
• MAPLE: Encoding Dexterous Robotic Manipulation Priors Learned From Egocentric Videos

Related

• Learning Dexterous In-Hand Manipulation
• Learning dexterous in-hand manipulation
• Rotating without Seeing: Towards In-hand Dexterity through Touch
• On the Feasibility of Learning Finger-gaiting In-hand Manipulation with Intrinsic Sensing
• A System for General In-Hand Object Re-Orientation