CAJ | 학술논문

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。
침대GPS접수궤관측조성분포적특점화입자려파존재적입자퇴화문제，제출기우유전산법화신경망락적중요성권치조정입자려파산법。장유전산법전국심우적수렴성여BP신경망락국부심우적쾌속성결합，개선입자적분포。장개진입자려파응용우GPS정위정도개선，여신경망락보조적입자려파진행비교，대비량자적유효입자수목화훈련오차。실측GPS수거처리험증표명，해산법증가료유효입자수목，유효억제료입자퇴화，재개선GPS정위정도상취득료량호적효과。
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 GANNWA-PF 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.