Deep Learning-based Indoor Positioning System Using Multiple Fingerprints

Indoor positioning system (IPS) based on Wi-Fi signal has gained increasing attentions during the past few years due to the low cost of infrastructure deployment. In the Wi-Fi signal based IPS, the channel state information (CSI) has been widely used as the feature of locations because the CSI signal is more stable and contains richer location-related information compared to the received signal strength indicator (RSSI). However, the performance of the IPS depending on a single access point (AP) can be much limited due to the multipath fading effect especially in most indoor environments involved with multiple non-line-of-sight (NLOS) propagation paths. In order to resolve this problem, in this paper, we propose a hybrid neural network that employs multiple APs to receive the CSI from. Each AP provides unique fingerprints to all the locations. By fully utilizing all the fingerprints gathered from the multiple APs, which reduces the NLOS effect, the robustness of the IPS is significantly improved.