Reinforcement Learning Based Power Allocation for 6G Heterogenous Networks

Heterogeneous networks (HetNets) play a crucial role in the context of 6G cellular networks, serving as a significant enabler for enhanced capacity and coverage. Nevertheless, the performance of multi-tiered architecture is impacted by interferences. While many strategies have been suggested to address interference management in HetNets and optimize power allocation, the challenge of concurrently ensuring quality of service (QoS) for both macro cell and small cell user equipment remains an ongoing area of study. The effectiveness of intelligent power distribution algorithms in HetNets has been shown by their inherent self-optimization abilities. In this paper, a power allocation strategy that is based on Reinforcement Learning (RL) is developed for relay-assisted HetNets. The proposed RL methodology aims to efficiently distribute power resources to the macro cell base station (MBS) and small cell base station (SBS) in order to satisfy the minimal capacity requirements of both macro cell user equipment (MUEs) and small cell user equipment (SUEs), hence ensuring the provision of adequate QoS. The RL algorithm under consideration maintains the minimum requirement of MUE and SUE along with a significant increase in their capacities. The modeling of a cellular network as a multi-agent network is achieved by attributing the role of an agent to each base station (BS). BS engage in interactions with adjacent BSs to facilitate the sharing of information and undertake self-optimization processes guided by an integrated rewards function. The simulation results demonstrate the efficiency and superiority of the proposed algorithm compared to the benchmark schemes.