Multiple Target Tracking in Urban Environments

We consider the problem of tracking a dynamically-varying and unknown number of targets in urban environments. Our proposed approach exploits available multipath measurements and directly incorporates them into a modified probability hypothesis density (PHD) filter to dynamically estimate both the number of targets and their corresponding parameters. The modified PHD incorporates a multipath-to-measurement association (MMA) scheme that adaptively estimates the best matched measurement return paths available at each time step. It takes into consideration that each target can generate multiple measurements due to multiple multipath returns. This approach avoids the use of the computationally intensive data association algorithm. It is also different from conventional multiple target tracking methods that first couple measurements to existing tracks through measurement-to-track associations and then estimate the target states using single target tracking techniques. The proposed algorithm is further generalized to more realistic urban terrain environments by including clutter as well as allowing for targets with varying kinematic models. Using simulations, we demonstrate the performance of the proposed approach in estimating both the number of targets and the corresponding target parameters at each time step.