Multi-Loco: Unifying Multi-Embodiment Legged Locomotion via Reinforcement Learning Augmented Diffusion

2025-06-13

AI Summary

Multi-morphological legged robots face significant challenges in generalizing locomotion policies due to discrepancies in observation/action spaces and dynamics across morphologies. Method: This paper proposes a morphology-agnostic diffusion-residual co-learning framework: (1) a generative diffusion model trained on cross-platform fused data to learn universal locomotion priors; and (2) a lightweight, shared residual reinforcement learning policy—built upon PPO—that enables morphology-specific action refinement and task adaptation. Contribution/Results: To our knowledge, this is the first framework enabling unified policy deployment across four heterogeneous legged platforms—including wheeled bipeds—with successful real-world transfer. Experiments demonstrate an average 10.35% improvement in simulated and physical-task returns; the wheeled biped achieves up to 13.57% gain. The approach significantly enhances robustness and cross-morphology generalization capability.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Generalizing locomotion policies across diverse legged robots with varying morphologies is a key challenge due to differences in observation/action dimensions and system dynamics. In this work, we propose Multi-Loco, a novel unified framework combining a morphology-agnostic generative diffusion model with a lightweight residual policy optimized via reinforcement learning (RL). The diffusion model captures morphology-invariant locomotion patterns from diverse cross-embodiment datasets, improving generalization and robustness. The residual policy is shared across all embodiments and refines the actions generated by the diffusion model, enhancing task-aware performance and robustness for real-world deployment. We evaluated our method with a rich library of four legged robots in both simulation and real-world experiments. Compared to a standard RL framework with PPO, our approach -- replacing the Gaussian policy with a diffusion model and residual term -- achieves a 10.35% average return improvement, with gains up to 13.57% in wheeled-biped locomotion tasks. These results highlight the benefits of cross-embodiment data and composite generative architectures in learning robust, generalized locomotion skills.

Problem

• Generalizing locomotion policies across diverse legged robots
• Handling varying morphologies and system dynamics
• Improving robustness and performance in real-world deployment

Innovation

• Combines diffusion model with RL policy
• Uses morphology-agnostic generative model
• Enhances robustness via residual policy