Absolute Positioning with Unsupervised Multipoint Channel Charting for 5G Networks

Abstract

The 5th generation mobile networks introduce large bandwidths with extended beamforming capabilities, which results in increased spatial selectivity of received channel state information. A channel chart is a map of the radio geometry that surrounds the base station and it can be generated in an unsupervised manner from the received channel state information without any knowledge of actual measurement locations. In this work, we generate channel charts for multiple base stations using multidimensional scaling and combine them for a better shape of the radio geometry. The combined chart cannot be directly applied for absolute positioning, but it can be extended. Extension is performed with affine and conformal mappings to the charts. The method for generating and combining the charts as well the method of extension to absolute positioning is explained. Evaluations are performed for two different scenarios, one of which is an open space scenario, and the other utilizes ray-tracing data. Finally, the charts are presented and analyzed together with positioning estimation results for the considered scenarios.