Performance bounds for tracking multiple objects using a single UAV

Abstract

In this paper we calculate probabilistic estimates for the size of an area a single unmanned aerial vehicle (UAV) can expect to monitor when tracking multiple objects. The objects are assumed to move according to a linear velocity model with Gaussian process noise. We use a Kalman filter to estimate the position of the objects. By using the covariance matrix of the Kalman filter, we can derive the necessary visitation period for a UAV to have a probability within a given confidence interval of redetecting the object at the estimated position. Then, we use this visitation period to calculate the probabilistic estimate for the area a single UAV can monitor. We demonstrate the results in Monte Carlo simulations.