Random Access Scheme for Machine Type Communication Networks Using Reinforcement Learning Approach

The random access (RA) procedure is pertinent to machine-type communication devices (MTCDs) effective utilization of cellular network resources. Although MTCDs produce a small amount of data that could be communicated during the random access process (RAP), a massive number of MTCDs attempting the RAP could cause congestion and network overload. MTCD growth for 5G networks and beyond will exacerbate these issues. Therefore, the lack of proper management of this growth on the cellular network will lead to an increase in latency and energy consumption for these devices. Data aggregation and reinforcement learning techniques like Q-learning provide a new auspicious strategy for solving machine-type communications (MTC) issues over cellular networks. Q-learning algorithms and data aggregation can tackle the RA problem. Thus, we integrate an effective data aggregation approach with reinforcement learning to meet MTCDs' Quality of Service (QoS) with lower energy usage. MTCDs are clustered and their data aggregated before using a Q-learning embedded random access strategy to adaptively address congestion and network overload. Our simulation showed that for massive MTCDs, our proposed technique outperformed the 3GPP access class barring (ACB) scheme in terms of delay and energy usage.