A modified federated Student’s t-based variational adaptive Kalman filter for multi-sensor information fusion

Abstract

To overcome the problem of poor performance of multi-sensor fusion algorithm in linear systems with time-varying noise and outliers, a modified federated Student’s t-based variational robust adaptive Kalman filter (MFSTVRAKF) is proposed in this article. In this algorithm, an improved Student’s t-based variational robust adaptive Kalman filter (ISTVRAKF) is developed for adaptively mitigating the effects of time-varying noise and outliers of each local sensor, which improves the estimation accuracy of the algorithm. It reorganizes and derives a simplified variational filter to approximate the prior distribution of the measurement noise covariance via the Student’s t distribution to eliminate outliers. Subsequently, an adaptive factor is introduced into ISTVRAKF for adjusting the weights of prediction and measurement to further exclude outliers. Moreover, a modified adaptive information sharing factor is devised to adjust the multi-sensor fusion weights in accordance with the performance of local filter, which improves the fusion accuracy of the MFSTVRAKF algorithm. Simulations exhibit that the proposed algorithm is capable of resisting the effects of time-varying noise and outliers and achieving higher accuracy compared to other filtering algorithms and fusion algorithms.