Overview of the motion trajectories datasets

Evaluating the performance of a motion prediction algorithm requires choosing appropriate testing scenarios and accuracy metrics, as well as studying the method’s robustness against various variables, such as the number of interacting agents or amount of maneuvering in the data. Depending on the application area, the testing scenario maybe an intersection, a highway, a pedestrian crossing, shared urban street with heterogeneous agents, a home environment, or a crowded public space. Existing datasets, summarized in this comparison