MACHINE PERCEPTION AND ROBOTICS GROUP

Dept. of Robotics Science and Technology,
Chubu University

Trajectory forecasting to generate plausible pedestrian trajectories in crowded scenes requires an understanding of human-human social interactions. Groups of pedestrians with the social norm move along similar trajectories, while groups of pedestrians with different norms make changes to their trajectories to avoid a collision. This paper introduces a group-based forecasting module for modeling inter- and intra-group interactions to enable an understanding of the social norm of humans for trajectory forecasting. In addition, group-based forecasting module takes the trajectory predicted by another prospection module as input to consider potential interactions with other groups in the future. In this way, our method models the complex group-level social interactions in crowded scenes through the attention mechanism and predicts socially plausible trajectories in accordance with each social norm. Comparisons we conducted with state-of-the-art forecasting methods show the effectiveness of our approach on three publicly available crowd datasets (ETH, UCY, and SDD). From experimental results, our network enables to predict plausible social trajectories by introducing two forecasting modules.