Adaptive systems of particle filters

We study systems of particle filters that track targets based on data acquired from a network of sensors. We build on our previous concept of symbiotic particle filtering and propose a system of particle filters, where each one of them explores a state space of minimal dimension. The number of particle filters in the system varies in that more particle filters may be added to the system, some may be removed, and some may be merged or split with time. The decision for changing the number of filters in the system depends on the estimated states of the targets that are being tracked and the locations of the sensors that sense them. We demonstrate the performance of the system by computer simulations and compare it with that of a standard particle filter.