An Out-of-Sample Extension for Wireless Multipoint Channel Charting

Abstract

Channel-charting (CC) is a machine learning technique for learning a multi-cell radio map, which can be used for cognitive radio-resource-management (RRM) problems. Each base-station (BS) extracts features from the channel-state-information samples (CSI) from transmissions of user-equipment (UE) at different unknown locations. The multi-path channel components are estimated and used to construct a dissimilarity matrix between CSI samples at each BS. A fusion center combines the dissimilarity matrices of all base-stations, performs dimensional reduction based on manifold learning, constructing a Multipoint-CC (MPCC). The MPCC is a two dimension map, where the spatial difference between any pair of UEs closely approximates the distance between the clustered features. MPCC provides a mapping for any given trained UE location. To use MPCC for cognitive RRM tasks, CSI measurements for new UEs would be acquired, and these UEs would be placed on the radio map. Repeating the MPCC procedure for out-of-sample CSI measurements is computationally expensive. For this, extensions of MPCC to out-of-sample UE CSIs are investigated in this paper, when Laplacian-Eigenmaps (LE) is used for dimensional reduction. Simulation results are used to show the merits of the proposed approach.