中文摘要：

无线网络节点定位技术给人员提供了安全保障,在研究传统无线定位的基础上,同时针对现有的基于神经网络定位算法的精度不高等问题,提出了一种新型的基于PSO-BP网络的定位算法.为了提高系统精度首先采用卡尔曼算法进行滤波处理,然后通过一种PSO-BP算法对BP网络初始权值和阈值进行优化,并对比现有的RSSI算法,分析不同算法的性能.BP神经网络权值的修正依赖于非线性梯度值,易形成局部极小值,同时学习次数较多.实验证明,改进的PSO-BP算法在误差反传调整权值的基础上,采用改进的PSO算法的学习机制修正权值,增加了BP算法收敛速度和全局收敛性,提高了BP网络的学习能力.

英文摘要：

On the basis of research on traditional wireless location,while not high for the existing positioning accuracy of neural network algorithm,we propose a new type of location-based algorithm PSO-BP network.In order to improve the accuracy of the system first Kalman filtering algorithm,and then by means of a PSO-BP algorithm BP initial weights and thresholds to optimize and compare existing RSSI algorithm to analyze the performance of different algorithms.Fixed BP neural network weights depends on the non-linear gradient value,easy to form a local minimum,while learning more frequently.Experimental results show that the improved PSO-BP algorithm based on the error back propagation adjustment weights on the use of learning mechanism correction weights improved PSO algorithm,an increase of BP algorithm convergence speed and global convergence,improved BP network learning ability.