Multiple target tracking using a narrowband RF sensor array

This paper describes a method for detecting and tracking multiple moving targets using a multistatic narrowband radar sensor array. We show that the appropriate model for the sensor array measurements is non-linear and the statistics are non-Gaussian. We use this model in conjunction with a Bayesian estimation algorithm which constructs a probability density on the number of targets and their states. The density is approximated using a novel hybrid discrete grid/particle filter. We evaluate the performance on a set of experimental data from a 6-channel array to detect and track two targets that cross, make sharp turns, and move nearly cross-radially to some pairs for a significant portion of the collect.