Eratosthenes: Deep generative modeling for indoor and outdoor positioning with fingerprinting methods
Fingerprinting methods of localization, which use machine learning, generally offer high localization accuracy, but on the other hand, suffer from the disadvantage of requiring a tedious and costly data collection phase. In this project, we propose to partly overcome this constraint by the introduction of generative modeling.
In the framework of the Eratosthenes project we will, firstly, investigate the concept of fingerprint augmentation through generative modeling. Achieving a reliable generation of fingerprints could have a great impact on the way fingerprinting positioning systems are designed, tuned and deployed.
Secondly, we will examine the potential benefits of encoding the fingerprints in the more compact representation of a latent space provided by variational autoencoders.
Finally, the most ambitious part of the Eratosthenes project is the incorporation of crowdsourced, unlabeled data (fingerprints at unknown positions), in order to strengthen the predictive capabilities of the positioning model, by using a Semi-Supervised Learning approach.