基于遗传算法和神经网络的改进粒子滤波的GPS定位数据处理  

Improved Particle Filter Algorithm Based on Genetic Algorithm and Neural Network for GPS Positioning Data Processing

作　　者：王尔申[1] 庞涛[1] 李兴凯[1] 张芝贤[1] 

机构地区：[1]沈阳航空航天大学电子信息工程学院,沈阳110136

出　　处：《电子器件》2015年第2期410-413,共4页Chinese Journal of Electron Devices

基　　金：国家自然科学基金项目(61101161);航空科学基金项目(2011ZC54010);辽宁省自然科学基金(联合基金)项目(2013024003)

摘　　要：针对GPS接收机观测噪声分布的特点和粒子滤波存在的粒子退化问题,提出基于遗传算法和神经网络的重要性权值调整粒子滤波算法。将遗传算法全局寻优的收敛性与BP神经网络局部寻优的快速性结合,改善粒子的分布。将改进粒子滤波应用于GPS定位精度改善,与神经网络辅助的粒子滤波进行比较,对比两者的有效粒子数目和训练误差。实测GPS数据处理验证表明,该算法增加了有效粒子数目,有效抑制了粒子退化,在改善GPS定位精度上取得了良好的效果。According to the measurement noise feature of GPS receiver and the degeneracy phenomenon problem in the particle filter,the genetic algorithm and BP neural network are applied in basic particle filter algorithm. A new approach based on genetic algorithm-optimized neural network-based importance weights adjustment particle filter（ GANNWA-PF） is proposed,which improves the posterior probability distribution of the particles. This GANNWAPF algorithm is applied in GPS positioning accuracy improvement. The experimental results based on the collected real GPS data are compared with the neural network aided particle filter,and compared with the effective number of particles and particle distribution. The experimental results indicated that the GANNWA-PF can increase the number of effective particle,and effectively solve the particle degeneracy phenomena,improve the accuracy of GPS positioning data processing.

关 键 词：全球定位系统 定位精度 粒子滤波 遗传算法 神经网络 

分 类 号：V241.6[航空宇航科学与技术—飞行器设计] TN967.1[电子电信—信号与信息处理]