2022/03/15 by Shreya Kadambi, Kadambi, Shreya, Arash Behboodi +11 · 2 citations
Earth and Planetary Sciences · Engineering · #FOS: Computer and information sciences #FOS: Electrical engineering #Indoor and Outdoor Localization Technologies #Information Theory (cs.IT) #Machine Learning (cs.LG) #Millimeter-Wave Propagation and Modeling #Precipitation Measurement and Analysis #Signal Processing (eess.SP) #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2203.08264
openalex publication_date 2022/03/15 · openalex created_date 2022/05/05 · openalex updated_date 2026/07/28
We present a neural network architecture for jointly learning user locations\nand environment mapping up to isometry, in an unsupervised way, from channel\nstate information (CSI) values with no location information. The model is based\non an encoder-decoder architecture. The encoder network maps CSI values to the\nuser location. The decoder network models the physics of propagation by\nparametrizing the environment using virtual anchors. It aims at reconstructing,\nfrom the encoder output and virtual anchor location, the set of time of flights\n(ToFs) that are extracted from CSI using super-resolution methods. The neural\nnetwork task is set prediction and is accordingly trained end-to-end. The\nproposed model learns an interpretable latent, i.e., user location, by just\nenforcing a physics-based decoder. It is shown that the proposed model achieves\nsub-meter accuracy on synthetic ray tracing based datasets with single anchor\nSISO setup while recovering the environment map up to 4cm median error in a 2D\nenvironment and 15cm in a 3D environment\n