In this project our goal is to infer the location of a mobile device in an indoor setting based on its received Wifi signals. Indoor localization of mobile devices is a key task in many commercial and emergency applications. The setting is different from the outdoor localization challenge, since an indoor Wifi signal is usually diffracted, reflected and received at multiple time intervals and directions. The consequence is that it is not possible to use triangularization as done in an outdoor setting.

In our work we approach this task via manifold learning. Specifically, we compute a map between the manifold of the Wifi signals, and the given shape of the indoor venue. The mapping is done by a small number of points that are used to compute a transfer function between two graphs that are based on the signals and the location of random points in the indoor venue.

 

Links to papers:

https://ieeexplore.ieee.org/abstract/document/6733301

http://proceedings.mlr.press/v54/moscovich17a.html

https://arxiv.org/abs/2202.03239

Link to code: