Dynamic Node Collaboration for Mobile Multi-Target Tracking in Two-Tier Wireless Camera Sensor Networks

We propose a dynamic node collaboration scheme and a track-to-estimate association scheme for the Mobile Multi-Target Tracking (MMTT) problem in the two-tier Wireless Camera Sensor Network (WCSN). We apply the Particle Multi-Bernoulli (PCBMeMBer) filtering algorithm to implement our proposed sensor collaboration scheme, which includes the Cluster Head (CH) selection scheme and the cluster member selection scheme. At each time step, every CH activates one of the cluster members, which has the best view of the tracked targets, to be the CH for the next time step. Each new CH collects the sensors located within its communication range and activates the ones collaborating to obtain more information to be its cluster members. Furthermore, we also implement a Gaussian-Mixture CBMeMBer (GMCBMeMBer) filtering algorithm to develop a track-to-estimate association scheme, which associates the target identities with the achieved multi-target states, which are collected in Random Finite Set (RFS). The simulation results evaluate our proposed dynamic node collaboration scheme especially when the target dynamics and/or measurement process are severely nonlinear. The simulation results also the improvements in the target-position estimate accuracy by using our proposed target-identification scheme.