Bayesian estimation of transition probabilities in hybrid systems via convex optimization

In practice, the transition probability matrix (TPM) in the approach to track a maneuvering target is often unknown. We propose a new method to estimate the optimal TPM according to the maximum a posteriori (MAP) or maximum likelihood (ML) criterion via convex optimization. We apply the proposed method to the nonlinear/non- Gaussian cases, where the interacting multiple model (IMM) particle filter (IMMPF) is employed to estimate the corresponding base state. Simulation results of tracking a maneuvering target show the efficacy of the proposed method with improved performance.