基于遗传算法优化的支持向量回归的室内定位算法

摘要

Indoor positioning based on the received signal strength index (RSSI) of zigbee has received more and more attention due to its low cost, low hardware power consumption and easy implementation. In order to improve the indoor positioning accuracy of zigbee technology and reduce the adverse effects of environmental factors, an indoor positioning method based on genetic algorithm optimization support vector regression is proposed. It is divided into two stages: offline collection and online prediction. Offline collection is used to establish the fingerprint database, and online prediction is according to the training model to predict position. Firstly, the all collected data is processed by Kalman filter, and then the penalty parameter radial basis function (RBF) kernel width and loss function variable of support vector regression are optimized by genetic algorithm (GA-SVR), so that the support vector regression reaches the best position prediction performance. The experimental result in the actual scene shows that compared with the particle swarm optimization support vector regression (PSO-SVR), grid search optimization support vector regression (GS-SVR), support vector regression (SVR) and weighted K-nearest neighbor (WKNN) algorithm, it has better positioning performance.%基于zigbee接收信号强度指示的室内定位由于成本低, 硬件功耗低, 易于实现而受到越来越多的关注.为了提高zigbee技术的室内定位精度, 减少环境因素的不利影响, 提出了一种遗传算法优化支持向量回归的室内定位方法.该算法分为离线采集和在线预测两个阶段, 离线采集进行指纹数据库的建立, 在线预测则根据训练模型进行位置预测.首先所有的采集数据通过卡尔曼滤波进行处理, 然后通过遗传算法优化支持向量回归 (GA-SVR) 的惩罚参数、径向基函数 (RBF) 核宽度和损失函数变量, 从而使支持向量回归达到最好的位置预测性能.在实际场景中的实验结果表明, 与粒子群优化支持向量回归 (PSO-SVR) 、网格搜索优化支持向量回归 (GS-SVR) 、支持向量回归 (SVR) 和加权K最近邻 (WKNN) 算法相比, 该算法具有较好的定位性能.