UMC: Unified Resilient Controller for Legged Robots with Joint Malfunctions

TL;DR

This paper introduces UMC, a two-stage, model-free framework that enhances legged robots' resilience to joint malfunctions by using masking mechanisms, significantly improving task performance across various damage scenarios.

Contribution

The paper presents a novel, model-free, two-stage training framework with masking for resilient legged robot control, addressing generalization issues of prior damage adaptation methods.

Findings

Improves task completion by 36% with transformers and 39% with MLPs.

Effectively adapts to eight types of damage scenarios.

Demonstrates robustness and adaptability in three locomotion tasks.

Abstract

Adaptation to unpredictable damages is crucial for autonomous legged robots, yet existing methods based on multi-policy or meta-learning frameworks face challenges like limited generalization and complex maintenance. To address this issue, we first analyze and summarize eight types of damage scenarios, including sensor failures and joint malfunctions. Then, we propose a novel, model-free, two-stage training framework, Unified Malfunction Controller (UMC), incorporating a masking mechanism to enhance damage resilience. Specifically, the model is initially trained with normal environments to ensure robust performance under standard conditions. In the second stage, we use masks to prevent the legged robot from relying on malfunctioning limbs, enabling adaptive gait and movement adjustments upon malfunction. Experimental results demonstrate that our approach improves the task completion capability by an average of 36% for the transformer and 39% for the MLP across three locomotion tasks. The source code and trained models will be made available to the public.