Research On Antenna Selection In Massive MIMO Uplinks

Massive multiple-input multiple-output(massive MIMO) with large antenna array configured in the base station can obtain a large number of spatial degrees of freedom. It uses the same time-frequency resources for multiple users where each user may access a plurality of antennas that leads to user connection density increment, as well as energy efficiency and system capacity improvement. Compared to long term evolution(LTE), the throughput of massive MIMO may obtain more than 10 times upgrade. Although massive MIMO has great potential for application, there are many challenges need to be addressed. Firstly, spatial correlation exists due to limited space isolation between antennas. Especially in the uplink, limited space isolation between antennas at base station leads to anisotropy which affects all the diversity branches. Secondly, imperfect channel state information(CSI) and complicated wireless channel fading environment introduce channel estimation error which significantly decreases the capacity gain of the system. Therefore influence factors such as antenna spatial correlation and imperfect channel estimation should be taken into account in massive MIMO system design.This paper focuses on the design of antenna selection scheme for massive MIMO uplink with antenna spatial correlation and imperfect channel estimation for single antenna multi-user and multiple antennas multi-user, respectively. It adopts optimization techniques to perform antenna selection by exploiting the sparsity of antenna selection matrices at the received end.(1) For the single antenna massive MU-MIMO uplink, two antenna selection schemes are designed: antenna selection based on the multi-user overall and on the multiple users individually. Based on the sparse recovery algorithm, three optimization models are proposed for each proposed scheme, including the uncorrelated antenna space model, the antenna spatial correlation model and the antenna spatial correlation and channel estimation error model. The solutions of these models are iteratively calculated by regularized orthogonal matching pursuit(ROMP) algorithm. The proposed schemes are compared with the antenna selection scheme calculated by OMP algorithm and the maximum ratio combining(MRC) full antenna selection scheme in the simulation to verify theoretical analysis and evaluate their performance.(2) For the multiple antennas massive MU-MIMO uplink, antenna selection which starts with a single user is classified into two types: antenna selection based on all users' antennas and on the multiple users' antennas individually. Two optimization models are proposed for each proposed scheme, including the uncorrelated antenna space and the case where both antenna spatial correlation and channel estimation error occur. The solutions are calculated by ROMP algorithm. The proposed schemes are compared with the maximum ratio combining(MRC) full antenna selection scheme in the simulation to verify theoretical analysis and evaluate their performance.