Adversarial Skill Learning for Robust Manipulation

TL;DR

This paper proposes an adversarial skill learning algorithm based on soft actor-critic to enhance the robustness of robotic manipulation policies against disturbances, validated through extensive simulation and real-world experiments.

Contribution

It introduces a novel adversarial training mechanism for robotic manipulation using SAC, improving policy robustness in disturbed environments.

Findings

Policies are robust to internal and external disturbances.

The algorithm performs well in both simulation and real-world tests.

Robust manipulation is achieved through adversarial skill learning.

Abstract

Deep reinforcement learning has made significant progress in robotic manipulation tasks and it works well in the ideal disturbance-free environment. However, in a real-world environment, both internal and external disturbances are inevitable, thus the performance of the trained policy will dramatically drop. To improve the robustness of the policy, we introduce the adversarial training mechanism to the robotic manipulation tasks in this paper, and an adversarial skill learning algorithm based on soft actor-critic (SAC) is proposed for robust manipulation. Extensive experiments are conducted to demonstrate that the learned policy is robust to internal and external disturbances. Additionally, the proposed algorithm is evaluated in both the simulation environment and on the real robotic platform.