Automatic self-reconstruction model for radio map in Wi-Fi fingerprinting

Abstract

The Wi-Fi fingerprinting is applied to the indoor positioning system (IPS) based on the radio map, which is the established received signal strength indication (RSSI) fingerprint database. However, since static radio maps built offline are vulnerable to environmental dynamics, interruptions occur in the online phase. To solve this problem, in this paper, we propose an Automatic Self-Reconstruction (ASR) model that combines Radio Encoding-based Deep Fingerprint Positioning (RE-DFP) and Radio Anomaly Detecting (RAD), which enables seamless positioning of Wi-Fi fingerprinting. The proposed model consists of an encoder-decoder-based RE-DFP network to improve positioning accuracy and data efficiency and a RAD network to analyze environmental dynamics according to APs. In the proposed RE-DFP, a CNN-based encoder is applied to increase efficiency by minimizing the dependence of the radio map, and the position is estimated through an LSTM-based decoder. RAD identifies the AP, which is the cause of the change in the radio map and excludes it from the online phase, preventing a decrease in positioning accuracy. Simultaneously, the proposed model collects the RSSI of the identified AP through crowdsourcing and reconstructs the radio map by itself, enabling a seamless positioning service in the online phase.