基于子区域切分与SSA-XGBoost的室内定位方法

Indoor Localization Method Based on Subspace Segmentation and SSA-XGBoost

在利用位置指纹进行实时室内定位时,由于多径效应、信号闭塞或无线AP本身不稳定,而影响最终的定位效果。对此,提出了一种基于子区域切分结合麻雀算法(Sparrow Search Algorithm,SSA)优化XGBoost的室内定位方法。离线训练阶段,利用改进的FCM(Fuzzy C-means)算法和区域相关系数指标将待定位区域划分为多个子域,通过AP优化为每个子域选择最优AP集合。针对XGBoost算法性能易受到初始参数问题影响,利用麻雀算法对XGBoost初始参数寻优得到相对较优的参数,并分别为各个子区域构建SSA-XGBoost定位模型。在线定位阶段,目标点通过匹配子区域的聚类中心得到所属子区域,最终利用该子区域的定位模型预测目标点的位置。与其他定位算法相比,所提算法平均误差分别减少14.7%、22.4%、37.1%,证明所提方法在实际环境中较其他算法具有更好的定位效果。

When using location fingerprint for real-time indoor positioning,redundant AP will occur due to multipath effect,signal blocking or unstable wireless AP itself,which will affect the final positioning effect.For this reason,an indoor location method is presented,which is based on subarea segmentation combined with sparrow search algorithm(SSA)optimization.In the offline training phase,the area to be located is divided into several subdomains by using the improved FCM(Fuzzy C-means)algorithm and the regional correlation coefficient index,and the optimal AP set is selected for each subdomain by AP optimization.Because the performance of the algorithm is susceptible to initial parameter problems,the sparrow search algorithm is used to optimize the initial parameters to obtain relatively optimal parameters,and the positioning model is built for each subarea.In the online positioning stage.the subareas of the target points are acquired by matching the cluster centers of the subareas,and finally the positioning model of the subareas is used to predict the location of the target points.Comparing with other locating algorithms,the average error of the proposed algorithm is reduced by 14.7%,22.4%,37.1%,respectively,proving that the proposed method has better locating effect in the actual environment than other algorithms.