Towards Robust Offline Reinforcement Learning under Diverse Data Corruption

arXiv:2310.12955v3 Announce Type: replace
Abstract: Offline reinforcement learning (RL) presents a promising approach for learning reinforced policies from offline datasets without the need for costly or unsafe interactions with the environment. However, datasets collected by humans in real-world environments are often noisy and may even be maliciously corrupted, which can significantly degrade the performance of offline RL. In this work, we first investigate the performance of current offline RL algorithms under comprehensive data corruption, including states, actions, rewards, and dynamics. …

abstract arxiv corruption cs.ai cs.lg data datasets diverse environment environments however humans interactions offline reinforcement reinforcement learning robust the environment type world