Analogue radio over fiber aided MIMO design for learning assisted adaptive C-RAN downlink

Cloud/Centralised radio access network (C-RAN) architecture is recognised as a strong candidate for the next generation wireless standards, potentially reducing the total cost of ownership (TCO). Spatial modulation (SM) is a cost-effective multiple-input-multiple-output (MIMO) solution, where only a single-RF chain is required for transmission. In this context, we propose an analogue radio over fiber (A-RoF) aided MIMO techniques for a learning assisted adaptive C-RAN system, where SM combined with space-time block coding (STBC) is optically processed using optical index in the central unit (CU) of the C-RAN, which also is capable of tuning the connected remote radio heads (RRHs). Furthermore, to improve the spectral efficiency, we invoke our proposed flexible C-RAN architecture for implementing learning assisted transceiver adaptation, where the number of RRHs connected to a single user and its modulation techniques employed are controlled using the K-nearest neighbourhood (KNN) algorithm. Simulation results show that the BER performance operating with A-RoF is just marginally degraded compared to that operating without A-RoF, while benefiting from the energy- and cost-efficient C-RAN design. Moreover, we show that the learning assisted adaptation is capable of outperforming the classic threshold-based adaptation in terms of the achievable rate