摘要
为了解决最小二乘法需要测量数据的先验信息来构造协方差矩阵的问题,提出了基于RBF神经网络的蜂窝无线定位算法。它融合了移动基站提供的AOA,TOA和TDOA测量值来实现移动台的定位,利用神经网络较快的学习特性和逼近任意非线性映射的能力,使其适用于复杂的多径环境。对基于RBF神经网络的定位系统性能进行了仿真,结果表明,基于REF网络的蜂窝无线定位算法消除了定位模糊和基站非理想分布对定位精度的影响,在小区半径小于2 km的情况下,系统的定位精度在125 m时准确率可达67%,在300 m时准确率可达95%。
In order to solve the problem how to construct covariance matrix that used prescient information of measurement data in least-square-method, a cellular location algorithm based on the RBF neural network is proposed. The measurement of AOA,TOA and TDOA provided by mobile base station is fused to locate mobile. The fast study and non-linear approach capacity of the neural network is made use of to apply in complicated multi-path environment. The location performance of the RBF neural network is simulated. The simulation results indicate that the uncertainty of location and the effect of bad basement position are avoided while the cellular localization algorithm based on the RBF neural network is used. When the radius of cellular is less than 2 km, its location within 125 meters is 67% of the time and within 300 meters is 95% of the time.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2008年第9期1798-1800,共3页
Systems Engineering and Electronics
基金
陕西省自然科学基金资助课题(2004F12)
关键词
最小二乘法
蜂窝系统
神经网络
非视距传播
least-square-method
cellular system
neural network
non-line-of-sight
