Multiple model filter with regularized UK state constraint

To solve the problem for maneuvering target tracking in the presence of model uncertainty, multi-likelihood regions and observation noise, we derive and present a multi-model filter with regular UK state constraints. Soft spatio-temproal constraint is modeled according to the truncation theory and incorporated into the routine of the state model prediction and model state nonlinear filtering. To deal with the problem of model attribute ambiguity caused by model inaccuracy, the state model probability is predicted by weighted Monte Carlo importance samples. To implement the soft constraints of the system, the objective function is optimized by adding the internal and external regular terms of the system structure, based on this, the effective observation information is tracked optimally via the unscented Kalman filtering of the modeled-state with multiple likelihood, and the multivariable nonlinear closed-loop system can be thereby stably controlled. The simulation experiment results demonstrate that, compared with the traditional interactive multi-model filters, the proposed filter has better ability to track maneuvering targets.