Efficient target tracking using mobile sensors

A mathematical model for tracking of a moving target by multiple mobile sensors in the framework of a partially observable Markov decision process is discussed. Applications include the use of a fleet of unmanned aerial vehicles for purposes such as search, surveillance, and target tracking. Computationally efficient approximate policies for controlling the mobile sensors are proposed, and a guarantee on their performance losses relative to that of the optimal policy is provided. Simulation results show that our proposed policies do perform close to the optimum for certain stationary models in which a mobile sensor can always move as fast as the target.