An Approach to Socially Compliant Leader Following for Mobile Robots

Mobile robots are envisioned to provide more and more services in a shared environment with humans. A wide range of such tasks demand that the robot follows a human leader, including robotic coworkers in factories, autonomous shopping carts or robotic wheelchairs that autonomously navigate next to an accompanying pedestrian. Many authors proposed follow-the-leader approaches for mobile robots, which have also been applied to the problem of following pedestrians. Most of these approaches use local control methods to keep the robot at the desired position. However, they typically do not incorporate information about the natural navigation behavior of humans, who strongly interact with their environment. In this paper, we propose a learned, predictive model of interactive navigation behavior that enables a mobile robot to predict the trajectory of its leader and to compute a far-sighted plan that keeps the robot at its desired relative position. Extensive experiments in simulation as well as with a real robotic wheelchair suggest that our method outperforms state-of-the-art methods for following a human leader in wide variety of situations.