Robust Adaptive Transition Probability Matrix in Interacting Multiple Model With Polynomial Functions and Feedback Structure

In conventional interacting multiple model (IMM) systems, the transition probability matrix (TPM) is predetermined using prior information. However, this fixed configuration can lead to errors in state estimation, which has led to research focused on adaptively adjusting the transition probabilities. While IMM with adaptive transition probability improves estimation accuracy, the robustness of TPM is not guaranteed in systems with more than two models when sub-models switch. To solve this problem, this paper proposes a polynomial function-based correction function and a feedback structure to adjust the transition probabilities. The polynomial function-based correction function increases the probability of sub-models that match the current situation or stabilizes probabilities by suppressing noise. The feedback structure employs a first-order infinite impulse response filter, combining current and previous information to adaptively adjust and prevent rapid declines in transition probabilities. The proposed algorithm is integrated into the IMM algorithm to enhance performance. Simulation for three-model system comparing the proposed algorithm with other adaptive IMM algorithms shows improved robustness of the transition probabilities and the state estimation accuracy.