LAUSR

A robot navigating in dense crowds should react to the motion of nearby pedestrians. However, it could lead to unsafe, inefficient, and illegible robot motions. This article presents an anticipative framework that predicts pedestrians intentions and their interactions in crowds, and the robot accordingly seeks an optimal trajectory based on the prediction. We propose: 1) a pedestrian motion model considering both pedestrian intention and interaction and 2) a multiobjective cost function considering real-time calculation, collision avoidance, quality of motion, and progress toward the goal along the trajectory. An interactive model predictive control framework is formulated to optimize the robot trajectory. The effectiveness of the proposed approach is evaluated in multiple simulation scenarios and a real experiment. It is demonstrated that the proposed approach generates safe, efficient, and legible robot behaviors in real time in dense crowds.