The authors describe an extension to the interacting multiple model (IMM) algorithm using a fuzzy logic framework. Each Kalman filter is considered to be locally valid, applicable over a region of the target acceleration space defined by the filter conditioning value. The validity of each Kalman filter is determined using fuzzy sets and an estimate of the target acceleration. With suitable choice of the fuzzy set overlap, only a subset of models close to the acceleration estimate need be evaluated. Simulation results are included for a typical manoeuvring target scenario which illustrate the effect of different forms of fuzzy sets. These confirm that the approach compares favourably with IMM in terms of both estimation accuracy and computational load.
展开▼