predicting trajectories of pedestrians is quintessential for autonomous
robots which share the same environment with humans. In order to effectively
and safely interact with humans, trajectory prediction needs to be both precise
and computationally efficient. In this work, we propose a