A Reinforcement Learning Approach for Network Slicing in 5G Networks

The emergence of the 5G ecosystem has revolutionized the landscape of communication networks, acting as a catalyst for digital transformation for individuals, companies, and industries. Efficient resource and slice management are vital in 5G networks to ensure the quality of service. To achieve this, a Reinforcement Learning (RL) approach is presented, which trains an intelligent agent to allocate slices in a 5G environment. Different RL algorithms such as Soft Actor Critic (SAC) and Deep Q-Networks (DQN) with some variants such as Double DQN, Dueling DQN, and Prioritized Experience Replay are used to optimize the allocation of slices based on the network state. The performance of the agent is compared with random allocation and heuristic-based methods. The objective is for the results to show that the proposed RL approach outperforms these methods, demonstrating the effectiveness of using RL for network slicing in 5G networks.