粒子滤波结合RBF神经网络用于室内定位  

Application of particle filter combined with RBF NN in indoor positioning

作　　者：李丽娜[1] 梁德骕 王越[1] 尤洪祥 

机构地区：[1]辽宁大学物理学院,辽宁沈阳110036 [2]中国联通系统集成有限公司辽宁省分公司,辽宁沈阳110036

出　　处：《计算机工程与设计》2017年第9期2509-2514,共6页Computer Engineering and Design

基　　金：国家自然科学基金青年基金项目(61403176);辽宁省教育厅科学技术研究基金项目(L2013003)

摘　　要：基于接收信号强度指示的室内定位方法在实际应用中定位精度不够理想,有待提高,鉴于此,提出一种改进的粒子滤波定位算法。将测距定位问题转化为非线性不相关方程组的最优化问题,根据测距误差大小对适应度值进行加权计算,平衡不同参考节点对定位目标的影响力,在一定程度上提高定位精度。提出利用RBF神经网络对室内传播损耗模型进行训练,进一步提高测距精度,保证定位优化问题模型的准确性。实验结果表明,所提定位算法平均定位误差约为30cm,基本可以满足一般的室内定位精度的要求。The positioning accuracy of indoor positioning methods based on the received signal strength indication is not ideal, which needed to be improved in the practical application. For the mentioned problems？ an improved particle filter algorithm was proposed. The ranging and positioning problem was converted into an optimization problem of nonlinear equations, and fitness value was processed through weighted calculation based on the ranging errors, the influences of different reference nodes on the target position were counterpoised,by which the positioning accuracy was improved to a certain extent. A new relevance vector machine algorithm based on mixed kernel functions was proposed to train the indoor propagation loss model, by which the ran-ging precision was improved and the accuracy of positioning optimization problem model was ensured. Experimental results show that the average positioning error of the proposed algorithm is about 30 cm, which can meet the indoor positioning accuracy re-quirements well.

关 键 词：室内定位 接收信号强度指示 粒子滤波 适应度加权计算 RBF神经网络 

分 类 号：TP391.44[自动化与计算机技术—计算机应用技术]