Deep reinforcement learning-based handover management for an advanced sliced vehicular network

Transport development impacts various aspects of life, including the economy, education, and healthcare. The concept of vehicle-to-everything (V2X) aims to provide safer and more satisfying travel experiences through Fifth-Generation (5G) of cellular network technology. Network Slicing (NS) is a key technology in the 5G architecture that addresses personalized Quality Of Service (QoS) requirements and resource management challenges. Handover (HO) is an essential technique in telecommunication and mobile communication systems. This technique maintains uninterrupted connectivity for users for mobile and vehicular networks. However, the increase in the number of HOs seeks to mitigate network performance. For that reason, in this study the purpose is to find a solution to improve network performance by decreasing the number of HOs. The application of Proximal Policy Optimization (PPO), a Deep Reinforcement Learning (DRL) algorithm, in a sliced vehicular network is the method to use in this research. We compared the simulation results obtained using PPO with a state-of-the-art algorithm and other DRL algorithms to identify the most effective approach. Our results show that PPO achieves superior performance, with approximately 80% and 20% improvements in cumulative reward and HO numbers, respectively. These findings indicate that PPO is the most suitable algorithm for sliced vehicular networks.