Adaptive Modified Transition Probability IMM Algorithm for Maneuvering Target Tracking
Gai Hang, Jianwei Wang · 2024
The conventional Interacting Multiple Model algorithm (IMM) adopts a static Markov Transition Probability Matrix (TPM), which lacks sensitivity to model switching during maneuvering phase, thus causing a degradation in tracking accuracy. To address this problem, this paper introduces a novel two-stage transition probability correction method to dynamically modify the elements of the TPM. In the first stage, the correction factor is constructed using the difference in model probabilities between adjacent time steps, and each element of the TPM is individually adjusted. Furthermore, an analysis of the correction factor values under various situations is conducted. In the second stage, to avoid potential switching delays caused by the correction in the first stage, a discriminant time window is set up to determine whether the target has maneuvered, thereby directly increasing the TPM corresponding to the new matching model to expedite the switching process. Simulation results verify that the proposed method can more accurately estimate the probability of matching models during different motion periods, resulting in higher target tracking accuracy compared to several existing adaptive algorithms.