Mixture-of-Experts RL for Fault-Tolerant Legged Locomotion

2026-06-24

AI Summary

This work addresses the insufficient locomotion robustness of legged robots under actuator failures and complex terrains by proposing a fault-aware modular control architecture. The approach uniquely integrates explicit fault diagnosis information with a Mixture-of-Experts (MoE) reinforcement learning framework, dynamically activating specialized control policies tailored to detected fault modes. This design achieves high performance while significantly reducing overall model capacity compared to monolithic alternatives. Experimental results demonstrate that the proposed method substantially outperforms single-policy baselines of comparable size across diverse actuator failure scenarios and maintains superior locomotion capabilities even under stringent computational constraints.

Abstract

Legged robots deployed in planetary exploration and other remote environments must maintain reliable locomotion despite actuator failures and challenging terrain conditions. Although reinforcement learning has achieved strong results in legged locomotion, monolithic policies can struggle to efficiently represent the diverse control strategies required to compensate for different fault conditions. In this work, we propose a fault-aware modular control architecture that explicitly leverages fault-diagnosis information to activate specialized control experts associated with distinct actuator failure modes. Experimental results show that explicit fault-conditioned modular policies consistently outperform monolithic policies of comparable size, achieving higher locomotion performance across failure scenarios. Moreover, the proposed modular architecture retains competitive performance even under significantly reduced network capacity, highlighting its suitability for compute-constrained robotic platforms, such as those typically employed in space applications. The code associated with this work is available at: https://github.com/iit-DLSLab/fault-locomotion-isaaclab.

Problem

• fault-tolerant locomotion
• legged robots
• actuator failures
• reinforcement learning
• modular control

Innovation

• Mixture-of-Experts
• fault-tolerant locomotion
• modular reinforcement learning
• actuator failure
• legged robots