Joint Allocation of Radio and Fronthaul Resources in Multi-Wavelength-Enabled C-RAN Based on Reinforcement Learning

Abstract

Cloud radio access network (C-RAN) is a key technology for the new-generation mobile network (5G). Multi-wavelength passive optical networks (PONs) such as wavelength division multiplexing (WDM) and time wavelength division multiplexing (TWDM) PONs are outstanding solutions for providing a sufficient bandwidth to support mobile fronthaul in C-RAN-based 5G architecture. In this paper, a joint allocation framework for multi-wavelength PONs fronthaul uplink resources and C-RAN radio interface uplink resources is presented. From the principle that the uplink resource allocation in mobile networks is an NP-hard optimization problem, this paper proposes a novel uplink scheduling algorithm based on reinforcement learning (RL). The performance of the algorithm is evaluated with numerical simulations and compared with two other algorithms from the literature; namely genetic algorithm (GA) and tabu search (TS) algorithm. The simulation results show that the new algorithm achieves higher system throughput, faster convergence and lower scheduling latency compared to the two other algorithms. Also, show that RL-based adaptive allocation of fronthaul transport block size based on actual radio resource block capacity can significantly reduce the capacity required on fronthaul link and decrease the total end to end uplink scheduling latency compared to the static allocation method.