基于自适应进化极端学习机的室内定位算法

摘要

为改进经典射频识别（RFID）室内定位算法LANDMARC与VIRE ，提出一种SEVIRE算法。为定位区域的接收信号强度值（RSSI）空间关联建模并用自适应进化极端学习机（SaE‐ELM ）离线训练，将在线采集的信号输入训练好的SaE‐ELM ，计算虚拟标签的RSSI值。在线定位时，为每个独立的阅读器寻找合适的阈值，引入 Q‐function减小信号波动对定位的影响，进一步提高定位精度。实验结果表明，在使用2.4 GHz频段阅读器的情况下，SEVIRE算法90％的定位误差在2 m以内，平均定位误差约为1.72 m ，整体定位性能优于LANDMARC算法和VIRE算法。%To improve LANDMARC and VIRE ,which are the classical radio frequency identification (RFID) indoor location al‐gorithms ,the method SEVIRE was proposed .The self‐adaptive evolutionary extreme learning machine (SaE‐ELM ) was adopted to learn the RSSI signal subspace correlation model in indoor location area ,then the virtual RSSI value with signals were compu‐ted in the online phase by the trained SaE‐ELM .During the online phase ,the accuracy was further improved by finding appropriate threshold for each reader independently and introducing Q‐function to reduce the impacts of signal fluctuations on positioning . Experimental results show that the proposed method with 2.4 GHz reader consistently enhances the precision of indoor localiza‐tion compared with the LANDMARC approach and the VIRE approach ,and the average error of the SEVIRE is about 1.72 m , 90% errors are less than 2 m .