KPCA/改进RBF神经网络辅助的GPS/UWB协同定位方法  

KPCA/Improved RBF neural network aided GPS/UWB cooperative positioning method

作　　者：孙伟[1] 曹红阳 SUN Wei;CAO Hongyang(School of Geomatics,Liaoning Technical University,Fuxin,Liaoning 123000,China)

机构地区：[1]辽宁工程技术大学测绘与地理科学学院,辽宁阜新123000

出　　处：《导航定位学报》2022年第6期112-121,共10页Journal of Navigation and Positioning

基　　金：2019辽宁省“兴辽英才计划”青年拔尖人才(XLYC1907064);2018年度辽宁省“百千万人才工程”人选科技活动资助项目(辽百千万立项【2019】45号);辽宁工程技术大学学科创新团队资助项目(LNTU20TD-06)。

摘　　要：针对车辆协同定位系统中,由于超宽带(UWB)信号中断导致协同定位解算精度下降的问题,提出基于核主成分分析(KPCA)/改进径向基(RBF)神经网络辅助的全球定位系统(GPS)/UWB紧组合协同定位方法。通过KPCA提取输入数据的非线性主成分进行降维处理,并利用改进K均值算法及遗传算法(GA)优化RBF神经元中心及连接权值等重要参数;当UWB信号中断时,利用训练好的神经网络对GPS/UWB紧组合系统进行补偿,解算出可靠的状态估计信息。实验结果表明,所提算法与无辅助时及基于粒子群(PSO)/RBF辅助时相比,平均定位误差分别减小56.1%和28%,有效提升车辆协同定位精度及稳定性。Aiming at the problem that the solution accuracy of cooperative positioning system is reduced due to interruption of ultra wide band(UWB) signal in vehicle cooperative positioning system, this paper proposed a tight combined global position system(GPS)/UWB cooperative positioning method assisted by kernel principal component analysis(KPCA) and improved radial basis function(RBF) neural network. KPCA is used to extract the nonlinear principal components of the input data and reduce the dimension. Meanwhile, improved K-means algorithm(K-means++) and genetic algorithm(GA) were used to optimize the important parameters such as RBF neuron center and connection weight. When the UWB signal is interrupted, the trained neural network is used to compensate the GPS/UWB tight combination system, calculate reliable state estimation information.Experimental result showed that the proposed algorithm reduces the average positioning error by 56.1% and 28%, respectively,compared with the unaided and particle swarm optimization(PSO)-RBF algorithm assisted, which effectively improves the accuracy and stability of vehicle cooperative positioning.

关 键 词：神经网络 紧组合 协同定位 核主成分分析 遗传算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]