NeoRL: A Near Real-World Benchmark for Offline Reinforcement Learning

TL;DR

NeoRL introduces a near real-world offline reinforcement learning benchmark with limited datasets and validation sets, addressing the gap between simulated benchmarks and real-world constraints, and evaluating existing algorithms in more realistic scenarios.

Contribution

The paper presents NeoRL, a new offline RL benchmark with controlled dataset sizes and validation sets, reflecting real-world constraints and providing a more practical evaluation environment.

Findings

Existing offline RL algorithms perform worse than deterministic policies on NeoRL.

Offline policy evaluation methods show limited effectiveness in the NeoRL setting.

NeoRL highlights the importance of realistic data collection and validation in offline RL.

Abstract

Offline reinforcement learning (RL) aims at learning a good policy from a batch of collected data, without extra interactions with the environment during training. However, current offline RL benchmarks commonly have a large reality gap, because they involve large datasets collected by highly exploratory policies, and the trained policy is directly evaluated in the environment. In real-world situations, running a highly exploratory policy is prohibited to ensure system safety, the data is commonly very limited, and a trained policy should be well validated before deployment. In this paper, we present a near real-world offline RL benchmark, named NeoRL, which contains datasets from various domains with controlled sizes, and extra test datasets for policy validation. We evaluate existing offline RL algorithms on NeoRL and argue that the performance of a policy should also be compared with the deterministic version of the behavior policy, instead of the dataset reward. The empirical results demonstrate that the tested offline RL algorithms become less competitive to the deterministic policy on many datasets, and the offline policy evaluation hardly helps. The NeoRL suit can be found at http://polixir.ai/research/neorl. We hope this work will shed some light on future research and draw more attention when deploying RL in real-world systems.