THOMAS: Trajectory Heatmap Output with learned Multi-Agent SamplingDownload PDF

Published: 28 Jan 2022, Last Modified: 13 Feb 2023ICLR 2022 PosterReaders: Everyone
Keywords: Trajectory prediction, Multi-agent, Motion forecasting, Motion estimation, Autonomous driving
Abstract: In this paper, we propose THOMAS, a joint multi-agent trajectory prediction framework allowing for an efficient and consistent prediction of multi-agent multi-modal trajectories. We present a unified model architecture for simultaneous agent future heatmap estimation, in which we leverage hierarchical and sparse image generation for fast and memory-efficient inference. We propose a learnable trajectory recombination model that takes as input a set of predicted trajectories for each agent and outputs its consistent reordered recombination. This recombination module is able to realign the initially independent modalities so that they do no collide and are coherent with each other. We report our results on the Interaction multi-agent prediction challenge and rank $1^{st}$ on the online test leaderboard.
One-sentence Summary: We propose a solution for multi-agent coherent multimodal trajectory prediction by learning a recombination of each agent predicted modalities.
17 Replies