基于无线信号K-M模型的三维室内定位算法研究

3D INDOOR POSITIONING ALGORITHM BASED ON WIRELESS SIGNAL K-M MODEL

针对无线信号在室内环境中易受到干扰、波动较大等问题,提出一种改进粒子群优化RBF神经网络的无线信号K-M传播测距模型。利用RBF的非线性特性模拟室内传播的复杂性,以信号接入节点(AP)发射功率、路径损耗因子、未知节点(RP)接收信号强度值RSS等构建模型,预测输出AP与RP之间的距离d。以d为半径,AP为球心,建立多个球体方程,采用极大似然(MLE)采样方程组与RSS-d加权质心混合定位算法,粗略估算未知节点位置信息,再利用加权质心法来进一步提高RP的定位精度。通过MATLAB实验仿真表明,与常见的优化算法对比,该模型预测距离误差更小,平均距离误差为1.3 m;RP的三轴坐标平均误差分别为x轴1.55 m、y轴1.48 m、z轴0.98 m,表明该模型提高了定位精度。

Aiming at the problems that wireless signals are susceptible to interference and fluctuation in indoor environment,we propose a wireless signal K-M propagation ranging model based on improved particle swarm optimization RBF neural network.The nonlinear characteristics of RBF were used to simulate the complexity of indoor propagation and the transmission power of signal access node(AP),path loss factor,and the received signal strength value(RSS)of unknown node(RP)were used to build the model to predict the distance d between the output AP and RP.Taking d as the radius and AP as the center of the ball,we established multiple spherical equations.The maximum likelihood(MLE)sampling equations and the RSS-d weighted centroid hybrid localization algorithms were adopted to roughly estimate the unknown node position information.Then,the weighted centroid method was used to further improve the positioning accuracy of the RP.Simulation results of MATLAB show that compared with the common optimization algorithm,our model has smaller prediction distance error and the average distance error is 1.3 m.The average error of the three axes of RP is 1.55 m for x axis,1.48 m for y axis and 0.98 m for z axis.It shows that the our model improves the positioning accuracy.