Intelligent Hybrid Scheduling Schemes in Device-Centric Communication for 5G Mobile Systems

Scheduling is the process of allocating resources for transmission of data. Among the various factors that establish when and what resources are assigned to a user, are the UE buffer status, QoS requirements, and channel conditions. Deep Learning and Reinforcement learning are techniques that can be utilized to improve scheduling. This paper includes a state-of-the-art review of the machine learning techniques and identifies gaps in the present schemes implemented. It was observed that scheme 4 which employed Multi-Agent Reinforcement Learning, achieved the optimal reward value faster as compared to the other schemes. Furthermore, two intelligent hybrid schemes are developed to improve the scheduling outcome in 5G mobile systems. The first hybrid scheme aims at enhancing throughput and reducing delay while ensuring fairness while the second hybrid scheme uses several agents for resource allocation while maximizing fairness and is expected to yield better results as compared to the former.