Global Navigation Satellite Systems (GNSS) such as GPS have been successfully deployed in the past two decades, and are indispensable for outdoor navigation.
However, people spend most of their times indoors, where limited or no GNSS service is available at all, because the satellites signals are too weak to penetrate the building. More importantly, GPS, with an average of 5-10 metre accuracy, cannot provide the indoor users with the positioning accuracy they need for room-level tracking.
My research utilises the existing structures (e.g. the WiFi network) and the smartphone to provide the positioning service.
The idea is creating a training database to reflect how the signal propagates inside the building. This is normally done by an expert holding a WiFi-enabled device (e.g. smart phone, laptop) walking around to record the signal at different locations, along with a positioning label. When the user wishes to discover her whereabouts, she measures the signal at her current position. The system then looks through the training database to find a nearest match. This process is called 'Location Fingerprinting'.
Nevertheless, the challenges for most signal-based indoor positioning systems are the unpredictable signal propagation caused by the complex building interiors, and the dynamic of the environment caused by the peoples' movements.
Hence, we proposed a novel technique to specify how certain the system's location estimation is. For example, at 90% confidence, our algorithm guarantees that the true location will be inside our prediction border up to 90% of the time.
Below is a video showcasing two versions of our algorithm estimating the user location in real time.
The blue dot is the true position. The Red dot is the system estimated position, whereas the black cross is the current state-of-the-art using Naïve Bayes.
In the top figure, the funky looking shape is the probability distribution of where the estimated locations are.
Positioning accuracy wise, both algorithms achieve below 2 metres most of the time, which are slightly better than Naïve Bayes. Range wise, regression version returns slightly smaller interval, at the same confidence level.
Another benefit of our approach is the ability to modify the size of our location estimation by altering the confidence level parameter. Below is another video demonstrating that effect.