EM-based Batch Algorithm for Trajectory Estimation of Multiple Targets

This study investigates the trajectory clustering problem in the presence of multiple moving targets which are approximated by straight motion over the observation interval and monitored by a radar sensor network. The assumption is that the nodes in the network lack the knowledge of the measurements’ association with each target, causing challenging estimation and classification tasks. To solve this problem, a novel unsupervised classifier based on expectation-maximization algorithm is devised by exploiting the latent variable model combined with cyclic estimation methods to obtain closed-form updates for the unknown parameters and build up a maximum a posteriori rule allowing for target measurement clustering. Finally, we analyze the effectiveness of the proposed algorithm through simulation experiments in a scenario with trajectory intersections.