Expert Dataset from CARLA

该数据集由一个使用密集点轨迹和经典PID控制器进行导航的确定性智能体构建，旨在为hGAIL架构提供标记的训练样本。该数据集包含十条预定义的轨迹，以每秒10赫兹的频率记录，每个剧集从随机点开始。密集轨迹提供高分辨率的点，而稀疏轨迹则指示大致方向。该数据集规模为21,287个训练样本（未压缩数据量为30GB），其任务是为城市环境中的自动驾驶导航提供支持。

This dataset is constructed by a deterministic agent that uses dense point trajectories and a classic PID controller for navigation, aiming to provide labeled training samples for the hGAIL architecture. The dataset includes ten predefined trajectories, recorded at a frequency of 10 Hz, with each episode starting from a random point. The dense trajectories provide high-resolution waypoints, while the sparse trajectories indicate the general driving direction. The dataset has 21,287 training samples with an uncompressed data volume of 30 GB, and it is intended to support autonomous driving navigation in urban environments.