GOSPA-Driven Multi-Bernoulli Gaussian Sensor Management

This paper presents a myopic sensor management algorithm for multiple target tracking with an agile single sensor that has a limited field of view (FOV), utilising a multi-Bernoulli filter. The sensor management algorithm aims to minimise the predicted mean square generalised optimal sub-pattern assignment (GOSPA) error. The aim is therefore to minimise the localisation error for detected targets, and the number of missed and false targets at the next time step. We present a tractable algorithm based on an upper bound of the mean square GOSPA error when single-target densities are Gaussian.