基于置信度加权的组合导航数据融合算法

 针对联邦滤波融合算法中由于模型量测噪声统计特性未能被准确描述导致其子滤波器误差变大，进而导致联邦滤波估计出现偏差的问题，为了改进联邦滤波融合方法，将模糊自适应卡尔曼滤波方法和置信度加权方法与联邦滤波融合方法相结合，应用于组合导航系统。该方法首先将模糊自适应卡尔曼滤波方法应用于各子滤波器，使其能够跟踪真实量测噪声统计特性。然后通过模糊方法计算得到各子滤波器的置信度，进而得到联邦滤波器的置信度，再由得到的置信度对各子滤波器及联邦滤波器输出进行加权，得到最终的全局输出。对车载组合导航系统的仿真结果表明，这种算法对量测噪声具有较强的自适应性，能够抑制置信度低的子滤波器在融合系统中所占的权重，提高联邦滤波融合算法的精度，是一种可行的车载组合导航数据融合算法。

Data Fusion of Integrated Navigation System Based on Confidence Weighted

A novel method on the data fusion framework is presented in this paper, taking the confidence of sub-filters into consideration. For the standard Kalman filter theory, the confidence of sub-filters is degraded by the time varied statistic of measurement noise, meanwhile the federated Kalman filter integrating these sub-filters is severely influenced. To cope with that, firstly, the fuzzy adaptive Kalman filter is applied to sub-filters to make them work optimally. Secondly, the confident function is used to compute the confidence of sub-filters, then the confidence of federated filter is calculated based on the confidence of sub-filters. And then the results of sub-filters and federated filter are weighted with the corresponding confidences. Simulations in INS/GPS/Odometer integrated navigation system demonstrate that the weight of sub-filter with low confidence is limited, and that the precision is improved as compared with the standard federated Kalman filter.