Abstract: Aiming at the problems of jitter and ineffective estimation of the unmeasured target state in the smooth variable structure filter, a time-varying smooth variable structure filter based on interactive multi-model is proposed. The algorithm initiates by estimating the target state through the smooth variable structure filter. Subsequently, it resolves the jitter problem by computing the time-varying smooth boundary layer and employing the tanh function instead of the saturation function to calculate the initial state gain. Following this, Bayesian formulas are applied to recompute the covariance matrix and state gain for updating the target state, effectively addressing the challenge of inefficiently estimating unmeasured states in the smooth variable structure filter. Lastly, the algorithm is integrated with the interactive multi-model approach to achieve effective tracking of maneuvering targets. Simulation results demonstrate that the proposed algorithm maintains its effectiveness in tracking maneuvering targets under conditions of model mismatch and changes in measurement noise. The simulation results indicate a significant improvement in tracking accuracy and enhanced robustness compared to typical target tracking methods.

Key words: smooth variable structure filter, time-varying smoothing boundary layer, interactive multi-model, state estimation, maneuvering target tracking