SURIMI: Supervised Radio Map Augmentation with Deep Learning and a Generative Adversarial Network for Fingerprint-based Indoor Positioning

Cloud-based Indoor Positioning Platform for Context-adaptivity in GNSS-denied Scenarios

“…This chapter has extended the results of the proposed Data Cleansing model that was published in [174], and the proposed Data Augmentation model that was published in [180].…”

“…% less than the CNNLoc on average. This chapter has extended the results of the proposed positioning model based on CNN-LSTM that was published in [180], and the proposed CNN-ELM model for Multi-building and Multi-oor classication that was published in [13]. This chapter covers the following points:…”

“…This dissertation extends some of the articles accepted in journals [32] and conferences [190,191,174,13,180], including two papers awarded as best student and runner-up papers. The rst award-winning paper was presented at ICUMT 2020 conference, where it won best student paper; it was extended in Chapter 5.…”

“…In this chapter, we introduce two ML models that exploit the advantages of DNN and ELM in order to provide a robust and generalised solution for ngerprintingbased indoor positioning applications. The rst model combines CNN and LSTM to estimate the user or device position (positioning model of the SURIMI framework introduced in Chapter 4) Quezada-Gaibor, Torres-Sospedra, Nurmi, Koucheryavy, and Huerta [180]. The second model uses CNN and ELM to provide a fast and accurate solution for classifying ngerprints into oor and building Quezada-Gaibor, Torres-Sospedra, Nurmi, Koucheryavy, and Huerta [13].…”

“…Training parameters for the cGAN using different methods and hyperparameters and the primary results. Reproduced with the permission from[180].…”