Pilot Assignment based on AoA Information using Channel Charting in Massive MIMO Systems

We consider a multi-point channel charting (CC) based uplink pilot allocation algorithm for a multi-cell network. We create the channel chart on an offline phase and map the base station (BS) locations to charting locations. The chart provides an approximation of the geometry of the network, with imprecise but indicative distances and directions. Accordingly we model the angle of arrival (AoA) of the signal arriving at the BS from a user equipment (UE) on the CC by a probability distribution. We make use of AoA distributions on the CC to construct similarity weights between users; users nearby in the angular domain have large similarity, whereas more distant users have smaller similarity values. The points of view of all BSs are considered to create a global similarity matrix. A constrained weighted graph colouring algorithm is then applied to allocate pilots. Simulation results show that the proposed pilot allocation algorithm outperforms state of the art CC-based pilot allocation algorithms. The performance of the CC-angle based pilot allocation scheme is comparable to that obtained by using known physical locations and angle spread information.