Humanoid Locomotion

Sim-to-real transfer for robust bipedal locomotion via inverse reinforcement learning and teacher-student training.

Overview

This project applies the Learn to Teach framework to the Digit humanoid robot, targeting sample-efficient sim-to-real transfer for robust locomotion across diverse terrains and conditions.

We use inverse reinforcement learning to infer reward functions from expert demonstrations, then train a teacher policy in simulation using a LiDAR-enriched observation space. The teacher’s knowledge is distilled into a deployable student policy that uses only proprioceptive sensing — enabling robust performance on hardware without privileged perception.

Publications

Links

• Project website
• arXiv: Learn to Teach
• arXiv: Infer and Adapt
• Video on YouTube