CMoE: Contrastive Mixture of Experts for Motion Control and Terrain Adaptation of Humanoid Robots

TL;DR

This paper introduces CMoE, a reinforcement learning framework that uses contrastive learning to improve terrain-specific expertise in humanoid robot motion control, enabling better navigation of complex terrains.

Contribution

The paper presents CMoE, a novel contrastive learning-based extension to the mixture of experts framework, enhancing expert specialization for terrain adaptation in humanoid robots.

Findings

Enables humanoid robots to traverse 20 cm high steps and 80 cm wide gaps.

Achieves robust, natural gait across diverse terrains.

Outperforms existing methods in complex terrain navigation.

Abstract

For effective deployment in real-world environments, humanoid robots must autonomously navigate a diverse range of complex terrains with abrupt transitions. While the Vanilla mixture of experts (MoE) framework is theoretically capable of modeling diverse terrain features, in practice, the gating network exhibits nearly uniform expert activations across different terrains, weakening the expert specialization and limiting the model's expressive power. To address this limitation, we introduce CMoE, a novel single-stage reinforcement learning framework that integrates contrastive learning to refine expert activation distributions. By imposing contrastive constraints, CMoE maximizes the consistency of expert activations within the same terrain while minimizing their similarity across different terrains, thereby encouraging experts to specialize in distinct terrain types. We validated our approach on the Unitree G1 humanoid robot through a series of challenging experiments. Results demonstrate that CMoE enables the robot to traverse continuous steps up to 20 cm high and gaps up to 80 cm wide, while achieving robust and natural gait across diverse mixed terrains, surpassing the limits of existing methods. To support further research and foster community development, we release our code publicly.