Abstract:

To improve the low tracking accuracy and solve the divergence problem in target tracking under the nonlinear and nonGaussian situation, an interacting multiple model algorithm is proposed based on cubature particle filter. The new algorithm utilizes cubature Kalman filter to incorporate the latest observation data and develop the importance density function, which is more close to the posterior density at the prior distribution updating phase, thus improving the particle filter performance. Simulation results show that, when compared with the interacting multiple model unscented particle filter, the proposed algorithm provides better filtering accuracy and stability while the average calculating time has no significant changes.