Humanoid Parkour Learning

TL;DR

This paper presents a novel end-to-end vision-based control framework enabling humanoid robots to perform complex parkour maneuvers autonomously without prior motion data, demonstrating versatile skills in diverse environments.

Contribution

It introduces a new learning framework for humanoid parkour that does not rely on motion priors and can transfer to mobile manipulation tasks.

Findings

Robot jumps on 0.42m platform

Leaps over 0.8m gaps and hurdles

Runs at 1.8m/s in outdoor environments

Abstract

Parkour is a grand challenge for legged locomotion, even for quadruped robots, requiring active perception and various maneuvers to overcome multiple challenging obstacles. Existing methods for humanoid locomotion either optimize a trajectory for a single parkour track or train a reinforcement learning policy only to walk with a significant amount of motion references. In this work, we propose a framework for learning an end-to-end vision-based whole-body-control parkour policy for humanoid robots that overcomes multiple parkour skills without any motion prior. Using the parkour policy, the humanoid robot can jump on a 0.42m platform, leap over hurdles, 0.8m gaps, and much more. It can also run at 1.8m/s in the wild and walk robustly on different terrains. We test our policy in indoor and outdoor environments to demonstrate that it can autonomously select parkour skills while following the rotation command of the joystick. We override the arm actions and show that this framework can easily transfer to humanoid mobile manipulation tasks. Videos can be found at https://humanoid4parkour.github.io