Decimeter Level Indoor Localisation with a Single WiFi Router using CSI Fingerprinting

Abstract

The accurate localisation of multiple objects and people in indoor environments is challenging. The problem is of immense importance in the context of Internet of Things (IoT) applications. In many scenarios, there is a high density of people and objects which need to be spatially and temporally tracked. As IoT applications find increasing use not only in industrial applications but also in novel areas such as healthcare in the hospital settings, indoor localisation challenges need redress. The paper addresses a novel technique for indoor localisation for a resolution of 10 cm for Line-of-sight (LoS), and Non-Line-of sight (NLoS) setup. The received signal strength indication (RSSI) based on WiFi signal is thoroughly studied in the past and has offered localisation solutions, yet it is highly sensitive to temporal and spatial variance due to multipath effect. The Channel state information (CSI) signal transmitted from WiFi hardware offers both space and time information, remains relatively stable with multipath propagation and interference, thereby the signal is considered highly suitable for designing localisation techniques with precision. The proposed solution is designed on an open source hardware, and CSI fingerprinting database for 100 locations that are spaced 10 cm apart for line-of-sight (LOS) and Non-line-of-sight (NLOS) configurations are acquired to generate classifier models using different Machine Learning algorithms. Random forest classifier model showed localisation results of 93.15% and 98.01% for LOS and NLOS respectively, for a resolution of 10 cm, which is reported for the first time. The design and technique can be further extended to various applications including patients localisation in dense waiting room of hospitals, home medical care, and other multiple tools localisation in an industrial environment using existing WiFi router infrastructure.