基于RNN的多传感器融合室内定位方法

针对室内空间内的人员定位困难问题进行了研究,提出了一种基于Wi-Fi指纹法和循环神经网络(re-current neural network,RNN)的多传感器融合室内定位算法.该算法将智能手机接收到的路由器信号强度作为时间序列输入RN N,通过RN N获得对行人精度较高的定位,与此同时获取智能手机中惯性测量单元提供的位置信息.随后,通过粒子滤波算法对两种定位方式的定位结果进行融合.在实际场景下设计了多组实验进行对比.实验结果表明,该算法定位平均误差为0.9 m,优于加权K近邻等算法,可以为行人提供实时的定位.

Multi-sensor fusion indoor localization method based on RNN

Aiming at the difficulty of locating people in indoor space, this paper proposed a multi-sensor fusion indoor location algorithm based on Wi-Fi fingerprint method and RNN. In this algorithm, it input the router signal strength received by the smart phone into RNN as time series to obtain the positioning of pedestrians with high accuracy through RNN. At the same time, it obtained the position information provided by the inertial measurement unit in the smart phone. Then, it fused the results of the two localization methods by particle filter algorithm. It designed several groups of experiments for comparison in the actual scene. Experimental results show that the average error of the proposed algorithm is 0.9 m, which is better than the weighted K-nearest neighbor algorithm and some other algorithms, and can provide real-time positioning for pedestrians.