Improving Single Reference Indoor Positioning Accuracy Through Machine Learning

Indoor Positioning Systems (IPSs) are able to provide information on the location of personnel and assets in various indoor environments such as offices, warehouses and factories. Using the position data from the IPS, Location Based Services (LBS) can be offered to personnel in offices. Also, various productivity improvement measures such as bottleneck identification, cutting wastage in searching for assets and floor layout optimization can be carried out in industry spaces like factories and warehouses. The widespread adoption of IPS is however currently hindered by insufficient positioning accuracy and the high infrastructure cost of installing and maintaining several reference nodes. This paper therefore proposes using just a single reference node and map matching for positioning along pre-determined routes. Furthermore, by employing Machine Learning, it is shown through simulation that the solution can be less complex and yet more accurate than traditional map matching, even in challenging indoor environments.