Particle Dance

该数据集来自粒子舞动领域，其状态和动作空间是连续的。在这个领域中，智能体学习跟随一个随机移动的粒子，同时保持一个安全距离。初始位置从区间[-1, 1]^4内均匀采样，初始速度固定为零。智能体根据与粒子的接近程度获得奖励，同时设有定义好的碰撞半径。任务是在安全约束下的强化学习。

This dataset originates from the particle dancing domain, which has continuous state and action spaces. In this domain, the agent learns to follow a randomly moving particle while maintaining a safe distance. Its initial positions are uniformly sampled from the interval [-1, 1]^4, and the initial velocity is fixed at zero. The agent receives rewards based on its proximity to the particle, with a well-defined collision radius set. The task is a reinforcement learning problem under safety constraints.