基于卡尔曼滤波的GPS导航系统故障检测

Fault detection in GPS navigation system based on Kalman filter

提出了一种将卡尔曼滤波器与神经网络相结合的导航系统故障检测结构.针对GPS系统的非线性特点,利用扩展卡尔曼滤波推算状态的预测值,并与实际观测值进行比较得到残差,同时采用Elman网络构成时延神经网络,利用残差的当前值及前多个历元的历史值进行故障检测.最后,计算机仿真验证表明,与χ2卡方检测方法比较,该结构不但可以快速检测出系统故障,提高故障检测的概率,而且能进行有效的故障隔离,具有很好的容错性能.

Combining the Kalman filter with neural networks, a new fault detection scheme is presented for navigation systems. Considering the nonlinear characters of the global positioning system （GPS）, an extra Kalman filter is used to calculate state prediction value. Then the error between prediction and measurement can be obtained. A time-delay neural network is constructed based on an Elman network, which makes use of the error and its history value to perform fault detection. The simulation results show that compared with X^2, the proposed algorithms can detect the fault of GPS, and have excellent fault tolerance performances, fast fault identifying and isolating ability.