TILoc: Improving the Robustness and Accuracy for Fingerprint-Based Indoor Localization

In WLAN fingerprint-based indoor localization, signal noise in the measurement of received signal strength indicator (RSSI) often results in matching a set of disperse reference points (RPs), leading to unsatisfactory estimation and weak robustness. To mitigate the noise problem, we propose a novel indoor positioning strategy, torus intersection localization (TILoc), aiming to improve the robustness and accuracy of fingerprint-based indoor localization. In the online phase, we design a new type of online RSSI fingerprints by filtering out unstable access points (APs). We use part of robust APs to construct RP torus and take the RPs in the intersection of RP tori as the nearest RPs. For reducing sparse spikes noise, we apply robust principal component analysis (RPCA) to train offline and online fingerprints. In addition, we take the AP's effect into consideration when we position a target. Our simulation and experiments show that the proposed algorithm outperforms other recent state-of-the-art algorithms in robustness and accuracy.