CMDB Mujoco Benchmark - Maze

该数据集是一个连续状态-动作空间任务，其中智能体必须在迷宫中导航以到达指定目标，同时避免碰撞。该数据集包含了复杂的几何形状和随机动力学，这导致了多样化的失败模式。它适用于高风险的安全关键环境，任务是在迷宫环境中进行强化学习。This dataset is a continuous state-action space task where an AI Agent must navigate a maze to reach a designated target while avoiding collisions. It features complex geometric shapes and stochastic dynamics, leading to diverse failure modes. It is tailored for high-stakes safety-critical environments, and is applicable to reinforcement learning tasks in maze environments.