神经网络优化5G定位结果域的分析与评估

Neural network optimization for the 5G positioning of the result domain

针对室内环境下的5G定位需求,提出了利用神经网络算法对粗略定位结果进行优化的方法,减小了多径、非视距传播造成的定位误差,改善了结果域的定位精度.优化算法利用测距定位中的到达时间(TOA)定位法和到达时间差(TDOA)定位法获得粗略定位结果,分别结合BP神经网络、Elman神经网络及通过遗传算法(GA)优化后的GA-BP神经网络、GA-Elman神经网络共利用4种神经网络进行训练,得到修正后的精确定位结果,并对4种神经网络算法进行了分析与评估.Elman算法相较于BP算法具有迭代收敛快、迭代次数少、误差改正好的特点,更适合5G定位结果域的优化;融入GA后结果精度均有所提高,其中GA-Elman算法能够训练得到最好的定位结果.

To meet the demand of the 5G positioning requirements in indoor environment,we proposed a method to optimize the rough positioning results by using neural network algorithms,which reduced the positioning error caused by multipath and non-line-of-sight propagation,and improved the positioning accuracy of the result domain.The optimization algorithm used the time of arrival(TOA)method and the time difference of arrival(TDOA)method in ranging positioning to obtain rough positioning results,and combined separately with BP neural network,Elman neural network,as well as genetic algorithm(GA)-BP network and GA-Elman network to obtain a better positioning results,then the four neural network algorithms were analyzed and evaluated.Compared with the BP algorithm,Elman algorithm has the characteristics of fast iteration convergence,few iterations and good error correction,which is more suitable for the optimization of the 5G localization result domain.The accuracy of the results is improved after incorporating the GA,among which the GA-Elman algorithm can be trained to obtain the best localization results.