Scheduling Variable Field-of-View Sensors for Tracking Multiple Objects

We propose the FOV of a sensor as a parameter to be controlled in real time to improve tracking performance. In contrast to earlier contributions, we take into account the measurement origin uncertainty in selecting the on-line sensor schedule. Two objects and two viewing modes are considered. The trace of the sum of approximations to the predicted covariances of the objects are used to select the mode at each instant of time; these approximations take into account the probability of target detection and the probability of false alarm. Simulations illustrate the superiority of the policy in comparison with fixed policies.