概率神经网络多历元残差犚犃犐犕算法

We propose a probabilistic neural network multi-epoch residual receiver autonomous integrity monitoring (RAIM) algorithm for civil aviation which can improve the detection capability of RAIM algorithm for fault deviation and reduce the minimum detectable deviation. A four-layer fault satellite detection model based on probabilistic neural network is constructed. Fault class and fault-free class training samples of pseudorange residuals are established using variance inflation model. The smoothing parameter of probabilistic neural network are optimized by particle swarm optimization algorithm to meet the false alarm rate. Thus, the similarity between the input multi-epoch pseudorange residual and the fault samples and fault-free samples can be calculated to determine whether the satellite is faulty. Simulation results suggest that optimizing the smoothing parameter can improve the fault detection ability of the proposed algorithm. Compared with the weighted least squares RAIM algorithm and the advanced RAIM (ARAIM) algorithm, the proposed algorithm can improve the detection performance of small pseudorange deviation and reduce the minimum detectable deviation under different fault conditions.