Safe and Robust Experience Sharing for Deterministic Policy Gradient Algorithms

TL;DR

This paper proposes a novel experience sharing mechanism with off-policy correction for deterministic policy gradient algorithms, enhancing safety and robustness in high-dimensional continuous tasks with limited replay memory.

Contribution

It introduces a simple, effective experience sharing method with off-policy correction that improves safety and robustness in deterministic policy gradient reinforcement learning.

Findings

Achieves safe experience sharing across multiple agents.

Exhibits robust performance with limited replay memory.

Outperforms baseline methods in continuous control tasks.

Abstract

Learning in high dimensional continuous tasks is challenging, mainly when the experience replay memory is very limited. We introduce a simple yet effective experience sharing mechanism for deterministic policies in continuous action domains for the future off-policy deep reinforcement learning applications in which the allocated memory for the experience replay buffer is limited. To overcome the extrapolation error induced by learning from other agents' experiences, we facilitate our algorithm with a novel off-policy correction technique without any action probability estimates. We test the effectiveness of our method in challenging OpenAI Gym continuous control tasks and conclude that it can achieve a safe experience sharing across multiple agents and exhibits a robust performance when the replay memory is strictly limited.