Indoor localization has become a hot topic in the last couple of years. It is particularly useful for large areas such as airport, malls or supermarkets. Its applications range from indoor navigation, targeted advertising to people and asset tracking. The localization problem is divided into two categories: outdoor and indoor. The global positioning system (GPS) is widely used for outdoor localization, especially for navigation. Unfortunately, this technique can’t be used for indoors because of the out of sight problem. Therefore, other technologies are adopted for indoor localization such as Bluetooth, vision, RFID, infrared, and Wi-Fi. In this project, we focus on Wi-Fi for the simple fact that we don’t need new equipment or a special setup since most indoor venues already contain access points. The two approaches used are fingerprinting with the received signal strength (RSSI) and trilateration using the round trip time (RTT) of the Wi-Fi signal.

The fingerprinting method consists of an offline phase and an online phase. During the offline phase, signal measurements are collected at various reference points in the indoor place. This is called a set of calibration data which will be used to construct a model. The calibration data usually consists of the RSSI measured from the different access points in the room. For the online phase, the model receives new signal measurements, which are independent of the calibration set, and will estimate the origin (position) of these measurements.

Google recently released a new update, Android P, which allows measuring the distance between a phone and its surrounding Wi-Fi access points using the RTT concept embedded in the IEEE 802.11mc standard. This protocol allows regular transfer of timestamps between an initiating STA (mobile) and a receiving STA (access point). This timestamp information makes it possible to calculate the RTT and to frequently estimate the distance between the two stations. Provided we have the RTT distance from at least three access points, a trilateration algorithm can be leveraged for building a user localization and occupancy detection.

As for the results, the fingerprinting method gives accurate distances within 2 meters from the ground truth. The results are impressive, however, environment dependent. We have to redo the offline data collection each time the environment changes which is unpractical. The geometric method however, gives us accurate distances between 1-2 meters after some data filtering and looks promising to develop an environment independent localization system.



Figure 1 represents an outdoor experiment done with only one access point to see the range in which RTT performs well. In conclusion, the ranging performs in the interval [1-3m] up to 40 meters and possibly further. Figure 2 represents a small 6x9 m2 room with 5 access points. The blue trace is the true position of the mobile and the green one is the predicted trace by the trilateration algorithm with the squared range least squared (SRLS) optimization.

Localization without GPS – a key issue for low-power IoT devices
A data driven approach for LoRa outdoor localization