Information-sharing approach to federated Kalman filtering

Summary form only given, as follows. An efficient information-sharing method is developed and applied to federated Kalman filters for distributed multisensor navigation systems. Based on a rigorous conservation-of-information principle, this novel method yields globally optimal or conservatively suboptimal filters with a variety of selectable operating characteristics. The method applies to decentralized systems in which one or more sensor-dedicated local filters feed a large master filter. The local filters operate in parallel, processing unique data from independent local sensors, and common data from a shared reference system. Data compression by the local filters further improves overall processing speed. The novel information-sharing technique allows the master filter to treat the local filter solutions as statistically independent, with no need to maintain local/local or local/master cross-correlation matrices. The method permits several modes of accumulated information sharing