Robot-Locomotion Tasks

该数据集包含了一系列机器人运动连续控制任务，其中包括“跑步任务”和“圆周任务”。在这些任务中，智能体被训练以在遵守安全限制的前提下最大化奖励。此外，该数据集为每项任务提供了多种机器人类型（如汽车、无人机、蚂蚁），旨在研究安全强化学习策略在面对观察性攻击时的鲁棒性。该数据集的任务主题是安全强化学习。

This dataset comprises a series of continuous robotic motion control tasks, including the "Running Task" and "Circular Task". In these tasks, AI agents are trained to maximize cumulative rewards while adhering to safety constraints. Furthermore, this dataset offers multiple robot platforms (e.g., car, drone, ant) for each task, with the objective of studying the robustness of safe reinforcement learning policies against observational attacks. The core research focus of the tasks within this dataset is safe reinforcement learning.