Multiple model tracker based on Gaussian mixture reduction for maneuvering targets in clutter
P.S. Maybeck, B.D. Smith · 2005
The problem of tracking targets in clutter naturally leads to a Gaussian mixture representation of the probability density function of the target state vector. Research reported previously reveals that a tracker using an integral square error (ISE) based mixture reduction algorithm can provide performance which is significantly better than any other known techniques using similar numbers of mixture components. One useful algorithm architecture for targets exhibiting very different trajectory characteristics over time would replace each Kalman filter within a conventional MMAE or IMM with an ISE-based algorithm that assumes the adequacy of the same particular dynamics model and discretized parameter choice ("mode"). The performance of such algorithms is evaluated, and compared to that of the corresponding MMAE or IMM based on Kalman filters in the same scenario except for being clutter-free.