AutoSceneGen Generated Dataset

该数据集是由AutoSceneGen框架生成的交通场景集合，包含了基于125个场景描述的264条行人轨迹和770条车辆轨迹。该数据集旨在通过提供多样化且关键安全的交通场景，来增强运动规划器的训练效果，其在轨迹预测任务上的表现优于现有的真实世界数据集。具体规模上，该数据集包含了264条行人轨迹（共1,848,000帧）和770条车辆轨迹（共15,400,000帧），其任务是为自动驾驶车辆进行轨迹预测。

This dataset is a collection of traffic scenarios generated by the AutoSceneGen framework, which includes 264 pedestrian trajectories and 770 vehicle trajectories derived from 125 scenario descriptions. It aims to improve the training of motion planners by offering diverse, safety-critical traffic scenarios, and it outperforms existing real-world datasets on trajectory prediction tasks. Specifically, this dataset contains 264 pedestrian trajectories (totaling 1,848,000 frames) and 770 vehicle trajectories (totaling 15,400,000 frames), and is designed for trajectory prediction tasks for autonomous vehicles.