A Combined GP-State Space Method for Efficient Crowd Mapping

Crowd sensing is an effective zero-cost method to map physical spatial fields by exploiting sensors already embedded in smartphones. The potentially huge amount of generated data and random measurement positions represent serious challenges to be addressed. In this paper we propose a combined Gaussian process (GP)-State space method for crowd mapping whose complexity and memory requirements for field representation do not depend on the number of data measured. The method is validated through an experimental campaign involving a high accuracy positioning system and a magnetic mobile sensor as data collector.