Robust Interactive Multiple Model based on Mahalanobis Distance for Target Tracking

This paper proposed a robust interactive multiple model (IMM) method based on Mahalanobis distance to deal with the issue of reduced filtering accuracy in target tracking due to interference from external environmental factors. Firstly, the nonlinear characteristics arising from the radar angle and range measurement models are addressed by integrating the Interactive Multiple Model (IMM) with the Cubature Kalman Filter (CKF). Then, the Mahalanobis distance is utilized to detect anomalous measurement data, and one of the sub-filters within the IMM is designed as a robust filter. Subsequently, a model set adjustment strategy is designed to enable the models to adaptively cover the actual measurement noise. The proposed IMM algorithm has a strong robustness to handle the varying the measurement noise. The simulation has been conducted to evaluate the efficacy of proposed robust IMM. The results demonstrate that it has excellent robustness compared to the traditional CKF, CKF-IMM and MRCKF.