Distributed Multimodel Bernoulli Filters for Maneuvering Target Tracking

We consider the distributed maneuvering target tracking in the wireless sensor network (WSN), where the maneuvering target may appear in or disappear from the surveillance region at random instances. We first develop a multi-model Bernoulli filter, which is based on the interacting multiple model method and the Bernoulli filter. Then, we present a centralized implementation (CI) of the multi-model Bernoulli filter, where a fictitious fusion center (FC) has the information of all sensors and is responsible for the CI. Finally, we present a consensus-based distributed implementation of the multi-model Bernoulli filter, where no FC exists in the WSN. Each sensor only communicates with its neighbors which enables the scalability in the network. By using average-consensus algorithms, each sensor iteratively diffuses its private information to the whole network. Numerical results illustrate that the distributed estimate at each sensor can be close to the global estimate at the FC.