WiFi Indoor Localization Based on Enhanced Dual-layer BiLSTM Network

With the popularity of IoT devices, the indoor Internet of Everything has put forward higher requirements for location-based services. And due to the shielding effect of buildings on GPS signals, GNSS is almost unavailable indoors. Therefore, it is of great significance to study indoor positioning technology with high reliability and accurate positioning. In recent years, the popularity of WiFi has promoted the development of indoor positioning technology, and WiFi-based positioning technology has become the focus of research. Existing research basically does not consider the impact of different APs on positioning results when establishing WiFi fingerprint databases, thus leading to large positioning errors. In order to solve the problem of low WiFi RSSI positioning accuracy, this paper proposes a neural network DL-BiLSTM based on an improved double-layer BiLSTM. The introduction of the Attention mechanism makes the network pay more attention to APs that are closer to the reference point, thus improving the positioning accuracy. Experimental results on public data sets show that the average error of our network is 0.95m with 100% of the errors under 2.5m. Compared with the existing method DLSTM, the mean position accuracy is improved by about 37%.