Robust-Gymnasium
收藏资源简介:
Robust-Gymnasium是一个开源的、用户友好的工具,专为评估和促进鲁棒强化学习算法的发展而设计。该数据集包含六十多个任务环境,跨越控制与机器人技术、安全强化学习和多智能体强化学习等领域,支持广泛的扰动类型,包括影响智能体观测状态、动作以及环境的扰动。这些任务环境通过整合不同类型、模式和频率的扰动器而构建,旨在为鲁棒强化学习算法提供全面的评估平台。
Robust-Gymnasium is an open-source, user-friendly tool specifically designed for evaluating and advancing the development of robust reinforcement learning algorithms. This dataset includes over sixty task environments spanning domains such as control and robotics, safe reinforcement learning, and multi-agent reinforcement learning, supporting a wide array of perturbation types including those affecting agent observations, actions, and the environment. These task environments are constructed by integrating perturbators of varying types, modalities and frequencies, with the goal of providing a comprehensive evaluation platform for robust reinforcement learning algorithms.

- 1Robust Gymnasium: A Unified Modular Benchmark for Robust Reinforcement Learning加州大学伯克利分校, 加州理工学院, 上海交通大学, 弗吉尼亚理工学院, 卡内基梅隆大学 · 2025年



