DeepDrive-Zero

该数据集是一个2D自动驾驶汽车模拟环境，使用自行车模型来模拟汽车行为，重点关注在没有信号灯的交叉路口场景中，两辆汽车如何在不发生碰撞的情况下进行合作。观察空间包含了关于智能体自身状态以及其他智能体状态的连续值，并且奖励函数基于多个参数设计。该数据集的规模为两个智能体，任务是在自动驾驶场景中进行多智能体协作。

This dataset is a 2D autonomous vehicle simulation environment that utilizes the bicycle model to simulate vehicle behaviors, with a primary focus on how two vehicles cooperate to avoid collisions at unsignalized intersections. The observation space contains continuous values corresponding to both the ego agent's own state and the states of other agents, and its reward function is designed based on multiple parameters. This dataset is configured for two agents, with the core task being multi-agent cooperation in autonomous driving scenarios.