Distributed Multisensor Fusion

Lucy Y. Pao Northwestern University Evanston, IL 60208 Abstract There have been several algorithms proposed for multisensor tracking of multiple objects using a centralized processing architecture, but because of considerations such as reliability, survivability, and communication bandwidth, distributed processing architectures are often the only alternative. The distributed fusion problem is more complex than the centralized fusion problem due to correlation across track estimates for the same object. We propose an approach for distributed sensor fusion that bypasses this correlation problem and allows previously developed centralized fusion methods to be employed after a measurement reconstruction procedure. Limitations and assumptions are discussed, and simulation results are presented to demonstrate the performances obtained with this approach.