Morphologically Symmetric Reinforcement Learning for Ambidextrous Bimanual Manipulation

2025-05-08

AI Summary

To address the challenges of role assignment, low sample efficiency, and poor generalization in bimanual dexterous manipulation for dual-arm robots, this paper proposes an equivariant cooperative control framework grounded in morphological symmetry. The method integrates equivariant reinforcement learning, hierarchical policy decomposition, and knowledge distillation. Its core contributions are: (i) the first formulation of robotic left-right morphological symmetry as an equivariance-inducing inductive bias within neural network architectures; and (ii) a three-stage paradigm—subtask decomposition, symmetric training, and policy distillation—that enables functional decoupling of hand roles and cross-arm experience sharing. Evaluated on six complex simulated tasks, the approach significantly outperforms established baselines. It is successfully deployed on two real-world dual-arm platforms and scales to quad-arm systems. Experiments demonstrate substantial improvements in both sample efficiency and cross-task generalization capability.

Technology Category

• Intelligent Robots: Manipulation
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Humans naturally exhibit bilateral symmetry in their gross manipulation skills, effortlessly mirroring simple actions between left and right hands. Bimanual robots-which also feature bilateral symmetry-should similarly exploit this property to perform tasks with either hand. Unlike humans, who often favor a dominant hand for fine dexterous skills, robots should ideally execute ambidextrous manipulation with equal proficiency. To this end, we introduce SYMDEX (SYMmetric DEXterity), a reinforcement learning framework for ambidextrous bi-manipulation that leverages the robot's inherent bilateral symmetry as an inductive bias. SYMDEX decomposes complex bimanual manipulation tasks into per-hand subtasks and trains dedicated policies for each. By exploiting bilateral symmetry via equivariant neural networks, experience from one arm is inherently leveraged by the opposite arm. We then distill the subtask policies into a global ambidextrous policy that is independent of the hand-task assignment. We evaluate SYMDEX on six challenging simulated manipulation tasks and demonstrate successful real-world deployment on two of them. Our approach strongly outperforms baselines on complex task in which the left and right hands perform different roles. We further demonstrate SYMDEX's scalability by extending it to a four-arm manipulation setup, where our symmetry-aware policies enable effective multi-arm collaboration and coordination. Our results highlight how structural symmetry as inductive bias in policy learning enhances sample efficiency, robustness, and generalization across diverse dexterous manipulation tasks.

Problem

• Enabling robots to perform ambidextrous bimanual manipulation tasks
• Leveraging bilateral symmetry for efficient policy learning
• Scaling symmetric policies to multi-arm robotic systems

Innovation

• Leverages bilateral symmetry via equivariant neural networks
• Decomposes tasks into per-hand subtasks with dedicated policies
• Distills subtask policies into global ambidextrous policy