Multi Sensor Information Fusion Algorithm Based on Smooth Variable Structure Filtering

Inaccurate modeling of noise statistical characteristics in civil aircraft navigation sensors can significantly degrade the overall positioning accuracy of the system. A decentralised multi-sensor information fusion algorithm based on smooth variable structure filter (SVSF) and performance degradation indication is proposed to address the above problem. The algorithm dynamically adjusts the information allocation coefficients of individual subsystems using a performance degradation factor derived from an optimally smoothed bounded layer. By integrating dynamic weight allocation and error covariance resetting, the proposed method effectively enhances the accuracy and robustness of navigation systems operating in complex environments. Experimental evaluations based on real flight trajectory data from the Beijing-Xi'an route demonstrate that, the proposed algorithm outperforms traditional Kalman filtering under constant and time-varying noise conditions. Results show a significant reduction in positioning errors caused by measurement outliers, validating the algorithm's effectiveness in mitigating sensor-induced disturbances.