Motion Planning for People Tracking in Uncertain and Dynamic Environments

Target tracking is an important capability for autonomous robots. The goal of this work is to construct motion strategies for a robot so that it can handle visual and mobility obstruction due to obstacles and maneuver effectively to track a mobile target in a dynamic, uncertain environment. There are two broad approaches to address dynamic changes and uncertainties in the environment: to react fast or to plan ahead. The choice often depends on the amount of prior information available on the environment and the target behavior. This paper gives an overview of our work on target tracking using these two approaches. First, we present a greedy algorithm. It uses purely local geometric information from the robot’s sensor to compute the robot’s motion at each time step, and yet carefully balances the robot’s ability to track the target in both the current and the future time. The algorithm uses only information from the robot’s sensor and requires no prior information on the environment or the target behavior. This has been shown to work well on a real robot with a 2-D laser sensor in a crowded school cafetaria. Second, we use partially observable Markov decision process (POMDP) to build a model of target behavior. As a result, the robot is capable of more sophisticated tracking behavior. For example, it may intentionally allow the target to get out of sight in order to minimize its own movement and save energy, but does not compromise long-term tracking performance. This is ongoing work and we show simulation results demonstrating the effectiveness of the approach.