Multi-BS association and Pilot Allocation via Pursuit Learning

Abstract

Pilot contamination (PC) interference causes an inaccurate user equipment’s (UE) channel estimations and significant signal-to-interference ratio (SINR) degradations. To combat the PC effect and to maximize network spectral efficiency, pilot allocation can be combined with multi-Base Station (BS) association and then solved by using learning algorithm efficiently. However, current methods separate the pilot allocation and multi-BS association in the network. This results in suboptimal network spectral efficiency performance and can cause an outage where some UEs are not allocated pilots due to the limited availability of pilots at each BS. In this paper, we propose a multi-BS association and pilot allocation optimization via pursuit learning. Here, we design a parallel pursuit learning algorithm that decomposes the optimization function into smaller entities called learning automata. Each learning automaton computes the joint pilot allocation and BS association solution in parallel, by using the reward from the environment. Simulation results show that our scheme outperforms the existing schemes and does not cause an outage.