OODD Benchmarks

OODD Benchmarks数据集由俄勒冈州立大学电气工程与计算机科学学院创建，专注于检测强化学习环境中的分布外动态变化。该数据集包含10000条正常轨迹和1000条异常轨迹，分别来自OpenAI Gym和Bullet物理环境的经典控制任务。数据集的创建过程涉及使用IQN和TD3算法训练控制策略，并在正常和异常环境下执行这些策略以生成轨迹。该数据集主要用于评估和改进在动态环境变化下的控制策略性能，特别是在强化学习和控制应用中。

The OODD Benchmarks dataset was created by the School of Electrical Engineering and Computer Science, Oregon State University, with a focus on detecting out-of-distribution dynamic changes in reinforcement learning environments. This dataset comprises 10,000 normal trajectories and 1,000 anomalous trajectories, derived from classic control tasks in OpenAI Gym and Bullet physics environments respectively. The development of this dataset involved training control policies using IQN and TD3 algorithms, then executing these policies in both normal and anomalous environments to generate the corresponding trajectories. This dataset is primarily used to evaluate and improve the performance of control policies under dynamic environment changes, especially in reinforcement learning and control-related applications.