Stealth Tracking of an Unpredictable Target among Obstacles

Abstract

This papers introduces the stealth tracking problem, in which a robot, equipped with visual sensors, tries to track a moving target among obstacles and, at the same time, remain hidden from the target. Obstacles impede both the tracker’s motion and visibility, but on the other hand, provide hiding places for the tracker. Our proposed tracking algorithm uses only local information from the tracker’s visual sensors and assumes no prior knowledge of target motion or a global map of the environment. It applies a local greedy strategy, and chooses the best move at each time step by minimizing the target’s escape risk, subject to the stealth constraint. The algorithm is efficient, taking O(n) time at each step, where n is the complexity of the tracker’s visibility region. Simulation experiments show that the algorithm performs well in difficult environments.