基于KDMSPCS-GRNN的室内定位技术研究  

Research on indoor positioning technology based on KDMSPCS-GRNN

作　　者：王超 单志勇[2] Wang Chao;Shan Zhiyong(School of Information Science and Technology,Donghua University,Shanghai 201620,China;Digital Textile Technology Ministry of Education Engineering Center,Shanghai 201620,China)

机构地区：[1]东华大学信息科学与技术学院,上海201620 [2]数字化纺织技术教育部工程中心,上海201620

出　　处：《信息技术与网络安全》2021年第4期20-27,45,共9页Information Technology and Network Security

基　　金：国家自然科学基金(61602110);教育部数字化纺织技术工程部资金(20D11426)。

摘　　要：针对利用广义神经网络(Generalized Regression Neural Network,GRNN)搭建的定位预测模型定位精度低、效率慢等问题,基于动态分群策略,提出一种线性递减粒子群(Linear Decreasing Contraction Particle Swarm Optimization,LDCPSO)和布谷鸟(Cuckoo Search,CS)混合寻优算法,并利用此算法为GRNN选择最优参数,构建定位预测模型。该算法主要利用K均值聚类算法(K-means)对整个种群进行周期性的分群,底层使用LDCPSO算法优化各个子群,并将最优粒子传至高层,高层使用CS算法优化各个子群的最优粒子,并将最终结果返回底层,执行下一次迭代。实验过程中,一方面将提出的算法应用于多个测试函数,结果表明该算法具有更好的收敛速度和收敛精度;另一方面利用该算法搭建定位模型,并与其他定位模型对比,结果显示该定位模型具有更好的定位效果。Aiming at the problems of low positioning accuracy and slow efficiency in the positioning prediction model built by the generalized neural network(GRNN),based on the dynamic clustering strategy,this paper proposed a Linear Decreasing Contraction Particle Swarm Optimization(LDCPSO)and Cuckoo Search(CS)hybrid optimization algorithm,and used this algorithm to select the optimal parameters for GRNN to construct a positioning prediction model.The algorithm mainly uses the K-means clustering algorithm to periodically group the entire population.The bottom layer uses the LDCPSO algorithm to optimize each subgroup,and the optimal particles are transmitted to the high level.The high level uses the CS algorithm to optimize the optimal particles of each subgroup and returns the final result to the bottom layer to execute the next iteration.During the experiment,on the one hand,the proposed algorithm was applied to multiple test functions,and the results showed that the algorithm has better convergence speed and accuracy;on the other hand,the algorithm was used to build a positioning model and compared with other positioning models,the results showed the positioning model has a better positioning effect.

关 键 词：LDCPSO算法 CS算法 K-mean算法 GRNN算法 测试函数 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]