NeoRL
收藏资源简介:
NeoRL是一个接近真实世界环境的离线强化学习基准数据集,由南京大学国家软件新技术重点实验室和Polixir Technologies共同创建。该数据集包含来自多个领域的数据,如机器人控制、工业控制、金融交易和城市管理,数据量最大可达9999条。NeoRL旨在通过提供受控大小的数据集和额外的测试数据集来验证策略,解决现有离线RL基准与真实世界应用之间的现实差距。数据集的应用领域广泛,旨在解决离线RL在实际系统部署中的安全、成本和伦理问题。
NeoRL is an offline reinforcement learning benchmark dataset that closely approximates real-world environments, co-created by the State Key Laboratory for Novel Software Technology at Nanjing University and Polixir Technologies. This dataset contains data from multiple domains including robot control, industrial control, financial trading and urban management, with a maximum of 9999 samples. NeoRL aims to verify policies by providing datasets with controlled scales and additional test datasets, so as to narrow the reality gap between existing offline RL benchmarks and real-world applications. The dataset has broad application scenarios, and is designed to address the safety, cost and ethical issues of offline RL during real-world system deployment.




