Student's t 滤波框架下的信息融合算法

Considering the problem of non-Gaussian measurement noises in multi-sensor system, an unscented quaternion filtering algorithm was designed based on Student's t distribution as a local filtering algorithm, by combining the Student's t filtering framework with the characteristics of quaternion and calculating the Student's t weighted integral function by unscented transformation. The optimal fusion weight coefficients were calculated by Lagrange multiplier method, and the local filtering results were fused by linear weighted fusion method. Quaternion was applied for simulation of the target attitude motion model. Three star sensors were used to observe the same target at the same time. The effectiveness of the proposed algorithm was verified through the comparison with the existing robust unscented Student's t filtering (RSTUF) algorithm. The simulation results show that the proposed distributed information fusion algorithm based on Student 's t filtering framework has higher estimation accuracy, convergence speed and numerical stability than RSTUF algorithm, which improves the estimation accuracy and fault tolerance of the algorithm by the complementarity of multi-observation information.