Recent channel state information (CSI)-based positioning pipelines rely on deep neural networks (DNNs) in order to learn a mapping from estimated CSI to position. Since real-world communication transceivers suffer from hardware impairments, CSI-based positioning systems typically rely on features that are designed by hand. In this paper, we propose a CSI-based positioning pipeline that directly takes raw CSI measurements and learns features using a structured DNN in order to generate probability maps describing the likelihood of the transmitter being at pre-defined grid points. To further improve the positioning accuracy of moving user equipments, we propose to fuse a time-series of learned CSI features or a timeseries of probability maps. To demonstrate the efficacy of our methods, we perform experiments with real-world indoor line-ofsight (LoS) and non-LoS channel measurements. We show that CSI feature learning and time-series fusion can reduce the mean distance error by up to 2.5× compared to the state-of-the-art.
I. INTRODUCTIONThe need for low-cost but accurate positioning systems is driven by recent trends in virtual reality, asset tracking, robotics, and industrial automation [2], [3]. Existing outdoor positioning solutions mostly rely on global navigation satellite systems (GNSS) that provide meter-level accuracy but require line-of-sight (LoS) satellite connectivity. High-precision indoor positioning solutions typically require specialized hardware that uses visible or infra-red light to localize objects with either active IR transmitting markers [4] or passive reflectors [1]. Such systems require unobstructed views, are affected by sunlight and reflective surfaces, and are costly.
A. CSI-Based Positioning with Neural NetworksLow-cost indoor positioning can be achieved with existing communication infrastructure that utilizes orthogonal fre-EG was with the School of Electrical and Computer Engineering,