Towards accurate indoor localization using iBeacons, fingerprinting and particle filtering

In this paper we present preliminary results of an experimental work aiming to exploit Bluetooth Low Energy technology for accurate indoor localization in highly structured environments. In the described experiment 31 iBeacons were placed in a small museum and their fingerprinting was modelled via progressively dense sampling, interpolation and Monte Carlo simulation. An accuracy of 2 meters was reached at an average density of about 0.3 samples square meter, and further increased with denser sampling in proximity of singularities of structural elements. Limits of two locate algorithms were analyzed. k-NN takes advantage of presence of near reference points but is fooled by minima of the distance function spatially apart from current location. Particle Filter takes advantage of accumulated evidence of locality but looses accuracy in areas where the distance field has low gradient values. Optimal beacons positioning is expected to improve accuracy of particle filtering in areas where the field is poorly structured.