An Adaptive Vehicle Tracking Enhancement Algorithm Based on Fuzzy Interacting Multiple Model Robust Cubature Kalman Filtering

Vehicle tracking is a core problem hindering multisensor fusion in intelligent driving. The interference caused by measurement outliers and motion estimation model mismatch seriously affects the estimation accuracy of target states. In view of these problems, an adaptive vehicle target tracking enhancement algorithm based on fuzzy interacting multiple model robust cubature Kalman filtering (FLIMM-IARCKF) is developed. In this algorithm, we constructed a derivative-free adaptive robust cubature Kalman filter (IARCKF) to suppress measurement outliers and errors in motion estimation modeling. Furthermore, a combined fuzzy reasoning method is developed to work with the interacting multiple model algorithm, which further enhances the target tracking performance by increasing the efficiency of model probability updating and by adaptively regulating the process noise covariance matrix. Simulated experiments verify the effectiveness of the IARCKF algorithm and reflect the advantages of the FLIMM-IARCKF algorithm in estimation accuracy, robustness, and model transformation efficiency when compared to the STCKF (Yun et al in Measurement 191:110063, 2022), Huber-CKF (Tseng et al in J Navig 70(3):527-546, 2017), MRCKF (Wu et al Acta Phys Sin 64(21):218401, 2015. https://doi.org/10.7498/aps.64.218401) and IMM-CKF (Song et al in ISA Trans 12(6):387-395, 2020) algorithms. The results show that the mean absolute percentage error (MAPE) of FLIMM-IARCKF in average position and velocity improved over the aforementioned approaches by 33.48%, and the average root-mean-square error (RMSE) improved by 32.64%. Real vehicle experiments showed that the average MAPE of FLIMM-IARCKF when used to determine position and velocity improved by 25.70%; the average RMSE improved by 28.96% when compared to the aforementioned algorithms, with an average operational time of 56.13 ms. Experimental results further revealed that FLIMM-IARCKF shows superior performance in vehicle target tracking without influencing execution efficiency.