Humanoid Locomotion Benchmark

该数据集提供了一个基准，比较了在训练人形机器人行走时，基础奖励、DRS（动态奖励塑造）条款以及PBRS（概率奖励塑造）条款之间的表现。此外，数据集还包含了展示不同奖励塑造方法学习性能的视频结果。每个案例（包括基础线、DRS和PBRS）都进行了十次学习运行，以评估不同方法在完成学习人形机器人行走任务时的表现。

This dataset provides a benchmark for comparing the performance of three approaches when training humanoid robots to walk: the basic reward, DRS (Dynamic Reward Shaping), and PBRS (Probabilistic Reward Shaping) methods. Additionally, the dataset contains video results demonstrating the learning performance of these different reward shaping approaches. Each of the three cases (including the baseline, DRS, and PBRS) underwent ten learning runs to evaluate the performance of each method when completing the humanoid robot walking training task.