Optimized Handover Decisions in 5G Mmwave UDNs Using Reinforcement Learning

The emergence of 5G and future technologies marks a significant advancement in cellular networks, driven by the growing demands for higher data rates and enhanced system capacity. The application of Ultra-Dense Network (UDN) is essential in meeting these demands. However, this massive deployment of small base stations introduces a major challenge in Handover procedures. With the increased number of base stations, handovers occur more frequently, causing not only increased signaling overhead but also a risk of degraded Quality of Experience (QoE) for users. In this paper, we present a handover decision scheme by implementing deep reinforcement learning (DRL) to reduce the frequency of unnecessary handovers in mmwave UDNs. The proposed model is based on Actor-Critic (A2C) algorithm, preferred for its effectiveness in optimizing complex and dynamic environments like UDN. Furthermore, Handover rate and signal strength is used as performance evaluation metrics. The results demonstrate that our proposed model successfully went through reinforcement training across various scenarios and eventually converges to an optimal policy.