Characteristic analysis and filtering algorithm design for UNGM model

The univariate non-stationary growth model (UNGM) is widely used in the verification of nonlinear filters, and unscented Kalman filter (UKF) often as reference for comparative analysis when using this to evaluate performance. However, due strong nonlinearity UNGM change properties with different parameter settings, estimation misalignment problem reasons will occur UKF filtering. To solve these problems, paper analyzes complex characteristics filtering process, proposes an sliding sampling module(SSUKF). algorithm optimized on basis UKF, can effectively deal by analyzing information process correcting distribution Sigma points real time. SSUKF applied under parameters compared bootstrap particle filter(BPF). simulation results show that UNGM, calculation speed better than BPF. Compared suitable a benchmark evaluating performance filters UNGM.