摘要
为了降低支持向量机(support vector machine,SVM)参数对定位精度的影响和降低其他优化算法优化参数易陷入局部最优解的风险,提出了一种基于人工蜂群算法(artificial bee colony,ABC)优化SVM的室内定位方法。在离线阶段通过射线跟踪技术生成RSSI仿真环境数据集,通过支持向量机构建RSSI与位置坐标的非线性映射模型,通过人工蜂群算法优化模型参数,得到最优模型;在线阶段,通过该模型对待定位点进行位置预测。首先将ABC-SVM定位方法与未经参数优化的SVM进行定位误差对比实验,结果表明,ABC-SVM在2 m内的定位精度达到86%,具有更高的定位精度;将ABC优化算法再与传统的ACO、PSO优化算法进行累积误差概率对比实验,结果表明,ABC-SVM算法在定位误差小于2 m的概率达到了90%。
In order to reduce the influence of support vector machine(SVM)parameters on the location accuracy,and mitigate the risk that the existing optimization algorithm and optimization parameters are easy to fall into the local optimal solution,an artificial bee colony algorithm(ABC)is proposed to optimize the indoor location method of SVM.In the offline phase,the RSSI simulation environment data set is generated by ray tracing technology,the nonlinear mapping model of RSSI and position coordinates is constructed by the support vector machine,the model parameters are optimized by the artificial bee colony algorithm,and finally the optimal model is obtained.In the online stage,the location point is predicted by using this model.Firstly,the comparison of the positioning error experiment between the ABC-SVM and the SVM without parameter optimization is carried out.The simulation results show that the positioning accuracy of ABC-SVM within 2 meters reaches 86%,which was higher than the accuracy rate of other methods.Secondly,the comparison of the cumulative error probability experiment between the traditional ACO and the PSO is conducted,the simulation results show that the probability of the ABC-SVM reaches 90%when the positioning error is less than 2 meters.
作者
孟国华
崔英花
MENG Guohua;CUI Yinghua(School of Information and Communication Engineering,Beijing Information Science&Technology University,Beijing 100192,China)
出处
《北京信息科技大学学报（自然科学版）》
2020年第2期33-37,共5页
Journal of Beijing Information Science and Technology University
基金
国家科学自然基金资助项目(61340005)
北京市自然科学基金面上项目(4132012)。
关键词
支持向量机
人工蜂群算法
室内定位
定位误差
累积误差概率
support vector machine
artificial bee colony
indoor location
positioning error
cumulative error probability
