MoRE: Mixture of Residual Experts for Humanoid Lifelike Gaits Learning on Complex Terrains

2025-06-10

AI Summary

Humanoid robots struggle to achieve controllable, human-like locomotion on complex, unstructured terrain. Method: This paper proposes MoRE—a Mixture-of-Residual-Experts reinforcement learning framework that fuses deep exteroceptive perception (depth camera) with high-level gait commands. It introduces a novel two-stage curriculum training paradigm and a multi-discriminator adversarial mechanism to decouple terrain adaptation from gait-style modulation. A tunable gait reward function enables real-time, smooth transitions among walking, jogging, and tiptoeing using only proprioceptive and depth-camera inputs. Results: Experiments demonstrate stable traversal of rubble, slopes, and stairs in both simulation and on a physical robot. Biomechanical metrics—including base height—closely match human kinematics. The approach significantly improves robustness and controllability of humanoid locomotion across diverse terrains.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Machine Learning: Mixture of Experts (MoE)

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Practical large-scale studies of user experience

Abstract

Humanoid robots have demonstrated robust locomotion capabilities using Reinforcement Learning (RL)-based approaches. Further, to obtain human-like behaviors, existing methods integrate human motion-tracking or motion prior in the RL framework. However, these methods are limited in flat terrains with proprioception only, restricting their abilities to traverse challenging terrains with human-like gaits. In this work, we propose a novel framework using a mixture of latent residual experts with multi-discriminators to train an RL policy, which is capable of traversing complex terrains in controllable lifelike gaits with exteroception. Our two-stage training pipeline first teaches the policy to traverse complex terrains using a depth camera, and then enables gait-commanded switching between human-like gait patterns. We also design gait rewards to adjust human-like behaviors like robot base height. Simulation and real-world experiments demonstrate that our framework exhibits exceptional performance in traversing complex terrains, and achieves seamless transitions between multiple human-like gait patterns.

Problem

• Enable humanoid robots to traverse complex terrains with lifelike gaits
• Overcome limitations of proprioception-only methods on flat terrains
• Achieve controllable gait switching between human-like patterns

Innovation

• Mixture of latent residual experts with multi-discriminators
• Two-stage training with depth camera and gait switching
• Gait rewards for adjusting human-like behaviors