An improved AIMM-UKF maneuvering target tracking algorithm

The adaptive Interactive multiple model (AIMM) tracking algorithm can adapt the Markov probability transition matrix, but when the target motion state is switched, it brings serious lag to the system model switching, resulting in the algorithm precision reduction. Aiming at the above problems, this paper proposes a model switching detection mechanism based on model likelihood function. When the algorithm detects that the system model is switched, it re-samples the corrected predicted value to generate Sigma points according to the AIMM-UKF filtering results, and then update the prediction error covariance matrix, cross-covariance matrix and new information covariance matrix. The Kalman filter gain is further adjusted to increase the effect of the measured value on the estimated value. Simulation results show that the proposed algorithm is $5.4 \%$ and $5.1 \%$ lower than the ARMSE of AIMM-UKF, and $13\%$ lower than the ARMSE of IMM-UKF, and has better filtering accuracy.