ABC优化BP神经网络算法在组合导航中的应用研究

摘要

The Back Propagation neural network based on Artificial Bee Colony algorithm is presented in this paper for integrated navigation. Firstly, on the condition that the BDS receiver receives signal normally, the SINS information such as the velocity and position is recognized as the input of ABC-based BP neural network and the output of this network is the output information of Kalman filter, so that the ABC-based BP neural network is trained and the corresponding math model is established. Then, in the case that the BDS receiver receives abnormal signal, the SINS information is recognized as the input of ABC-based BP neural network, and the established model is used to predict the output adjustment information, which makes the SINS adjusted. Finally, simulation results indicate that compared with the traditional BP neural network, the ABC-based BP neural network has better performance on positioning accuracy.%针对北斗/捷联惯导组合导航,提出一种基于人工蜂群ABC(Artificial Bee Colony)算法的反向传播BP(Back Propagation)神经网络算法.首先,在北斗卫星导航系统接收机正常接收信号时,将捷联惯性导航解算信息(速度、位置)作为网络输入,卡尔曼滤波输出信息(速度、位置校正量)作为网络输出,对ABCBP神经网络进行在线训练,建立ABCBP神经网络的映射数学模型.然后,在北斗卫星导航系统接收机信号失效情况下,将惯性导航解算信息作为网络输入,利用建立好的ABCBP神经网络预测输出校正量信息,以此来校正捷联惯导系统SINS(Strapdown Inertial Navigation System).最后,通过飞机飞行半物理仿真实验验证该算法的性能.仿真结果表明,ABCBP神经网络算法在定位精度方面具有更加优越的性能.