Rethinking Robustness Assessment: Adversarial Attacks on Learning-based Quadrupedal Locomotion Controllers

TL;DR

This paper introduces a computational method using sequential adversarial attacks to identify vulnerabilities in learning-based quadrupedal locomotion controllers, revealing potential weaknesses even in robust policies and aiding in their improvement.

Contribution

It presents a novel approach for quantitatively assessing the robustness of neural network-based locomotion controllers through adversarial sequences.

Findings

State-of-the-art controllers can fail under low-magnitude adversarial sequences.

The method is effective in simulation and real robot experiments.

Results can be used to enhance policy robustness and safety.

Abstract

Legged locomotion has recently achieved remarkable success with the progress of machine learning techniques, especially deep reinforcement learning (RL). Controllers employing neural networks have demonstrated empirical and qualitative robustness against real-world uncertainties, including sensor noise and external perturbations. However, formally investigating the vulnerabilities of these locomotion controllers remains a challenge. This difficulty arises from the requirement to pinpoint vulnerabilities across a long-tailed distribution within a high-dimensional, temporally sequential space. As a first step towards quantitative verification, we propose a computational method that leverages sequential adversarial attacks to identify weaknesses in learned locomotion controllers. Our research demonstrates that, even state-of-the-art robust controllers can fail significantly under well-designed, low-magnitude adversarial sequence. Through experiments in simulation and on the real robot, we validate our approach's effectiveness, and we illustrate how the results it generates can be used to robustify the original policy and offer valuable insights into the safety of these black-box policies. Project page: https://fanshi14.github.io/me/rss24.html