Multi-agent motion planning has recently attracted much interest in the research community and has many applications including robot navigation among pedestrians, self-driving cars, and drone shows. This problem: 1) concerns finding a one-to-one assignment between a set of given agents’ initial locations and the corresponding destination locations and 2) finding the continuous trajectories connecting each agent’s initial position to its goal location.

Since the past few decades, human trajectory forecasting has been a field of active research owing to its numerous real-world applications: evacuation situation analysis, deployment of intelligent transport systems, traffic operations, to name a few. We can cast the problem of human trajectory forecasting as learning a representation of human social interactions. However, social interactions in crowded environments are not only diverse but often subtle.