On a Problem of Multisensor Tracking

This paper describes a technique for combining state estimates from two nonlinear filters for the purpose of providing a global estimate that converges to the correct state value, while the estimates from the individual filters may not. The practical situation considered-occurs in target motion analysis. In this case, there are two extended Kalman filters in the tracking system, one processing bearing, and the other processing conical angle measurements provided by spatially separated sensors. Prior to an observer maneuver, the target parameters from each filter will not converge. However, by using the proposed technique, the global estimate-a combination of the two estimates--converges to the correct value without an observer maneuver, as the numerical results show.