Real-World Reinforcement Learning (RWRL) suite

该数据集名为RWRL套件，包含多个用于强化学习实验的领域，特别关注那些具有可解与不可解约束的任务。该套件专为软约束强化学习而设计，支持在不同约束违规阈值和性能指标下进行实验。其规模涵盖多个领域，包括倒立摆、步行者、四足动物和人形机器人等。任务类型为带有约束的强化学习。

This dataset, named the RWRL Suite, includes multiple domains for reinforcement learning experiments, with a particular focus on tasks with both solvable and unsolvable constraints. Specifically designed for soft constraint reinforcement learning, the suite supports experiments under varying constraint violation thresholds and performance metrics. It covers a diverse set of task domains, such as inverted pendulum, walker, quadruped, and humanoid robot, among others. The task type of this suite is constrained reinforcement learning.