How Soon Will Channel Charting Be Inapplicable in User Moving Scenarios?: Some Answers.

Channel Charting (CC) is a newly developing framework, which aims to learn the low dimensional representation of radio geometry from the channel state information (CSI) collected in the region of interest. But one key question, how soon will channel charting be inapplicable in user moving scenarios, even in certain setting? remains open. In this paper we present some answers by designing corresponding experiments with single cell and multi cell scenarios in LOS channel and multi-path channel settings. Our experimental results suggest that, in the communication scheme with user moving speed = 10 m/s, which is much faster than human walking, the constructed channel charting can provide up to 30 seconds support for future channel estimation. Our work shows that, machine learning and wireless big data based wireless channel learning may support complex communication tasks, especially for massive MIMO system in future communication systems.