基于GA-BP神经网络的多无人机协同目标定位  

Multi UAVs cooperative target localization based on GA-BP neural network

作　　者：彭俊澄 张薇[1] PENG Juncheng;ZHANG Wei(College of Information and Communication Engineering,Harbin Engineering University,Harbin 150001,China)

机构地区：[1]哈尔滨工程大学信息与通信工程学院,黑龙江哈尔滨150001

出　　处：《应用科技》2024年第3期141-149,共9页Applied Science and Technology

基　　金：国家重点实验室开放课题项目(CEMEE2021K0101A).

摘　　要：无人机作为高机动平台,往往存在较大的自身位置误差。针对该情况下特定区域内的辐射源目标,提出了一种遗传算法优化反向传播(genetic algorithm optimized back-propagation,GA-BP)神经网络的实时定位方法。首先从特定区域中获取不同已知目标的位置信息作为网络的期望输出,并计算它们的到达时间差作为网络的输入,构建训练数据集;然后利用遗传算法的自适应性,优化BP神经网络初始权重和阈值,使其能够快速跳出局部最优解,实现高精度定位。通过训练得到相应的GA-BP网络模型,未知目标可以通过该模型进行实时定位。仿真实验将该算法与两步加权最小二乘算法以及BP神经网络的定位结果进行对比,结果表明所提的方法定位精度更接近克拉美罗界。As highly mobile platforms,UAVs often have a large error in their own position.A genetic algorithm optimized back-propagation neural network(GA-BP)real-time localization method is proposed for multiple radiation source targets in a specific region in this contribution.Firstly,the position information of different known targets is obtained as the expected output of the network from a specific region and their arrival time differences are calculated as the input of the network to construct training data set;then,the initial weights and thresholds of the BP neural network are optimized by using the adaptive nature of the genetic algorithm so that it can quickly jump out of the local optimal solution to achieve high accuracy localization.The corresponding GA-BP network model is obtained through training,and the unknown target can be localized in real time by using this model.Simulation results show that,compared with the two-step weighted least squares algorithm and the BP neural network,the localization accuracy of the method suggested in this work is closer to the Cramer-Rao boundary.

关 键 词：无源定位 到达时间差 无人机 GA-BP神经网络 实时定位 克拉美罗界 多目标定位 站址误差 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]