HACL: History-Aware Curriculum Learning for Fast Locomotion

TL;DR

This paper introduces HACL, a history-aware curriculum learning algorithm that improves the speed and stability of quadrupedal and bipedal robot locomotion by leveraging past information through an RNN, demonstrating significant performance gains.

Contribution

The paper presents a novel history-aware curriculum learning method using RNNs to enhance robot locomotion speed and stability, validated on multiple robots in simulation and real-world.

Findings

Achieves peak forward velocity of 6.7 m/s on robots.

Outperforms prior algorithms by nearly 20%.

Effective in both simulated and real-world environments.

Abstract

We address the problem of agile and rapid locomotion, a key characteristic of quadrupedal and bipedal robots. We present a new algorithm that maintains stability and generates high-speed trajectories by considering the temporal aspect of locomotion. Our formulation takes into account past information based on a novel history-aware curriculum Learning (HACL) algorithm. We model the history of joint velocity commands with respect to the observed linear and angular rewards using a recurrent neural net (RNN). The hidden state helps the curriculum learn the relationship between the forward linear velocity and angular velocity commands and the rewards over a given time-step. We validate our approach on the MIT Mini Cheetah,Unitree Go1, and Go2 robots in a simulated environment and on a Unitree Go1 robot in real-world scenarios. In practice, HACL achieves peak forward velocity of 6.7 m/s for a given command velocity of 7m/s and outperforms prior locomotion algorithms by nearly 20%.