Evaluating the influence of parameter variations on multi-sensor tracking

In this paper we evaluate the influence of variations in the input parameters on the output of a multi-sensor tracking algorithm, using simulated data. The tracking algorithm is a classical Kalman filter using a probabilistic data association. The input to the tracker consists of contact files, each file containing all contacts identified for a specific per source / receiver / ping triplet. The input parameters that are varied are: 1) detection threshold used to identify the contacts, 2) ping repetition rate, 3) amplitude of contact position errors, 4) number of sensors used, 5) target signal-to-noise ratio, 6) relative ping time of sensors, and 7) waveform. A dasiastandardpsila set of tracker performance metrics is used to evaluate the tracker output and to look for trends in this output versus parameter values.