Deep Reinforcement Learning based traffic routing on Mobile Edge Networks

A potential paradigm that makes compute and storage resources available at edge of network is Mobile Edge Computing (MEC). In this paper, we propose a traffic routing algorithm for MEC networks leveraging Deep Reinforcement Learning (DRL). Our approach leverages a dataset of MEC network topologies and specifications to model the network and trains a DRL agent to select the optimal routing path on grounds of current state of the network. We adopt a Deep Q-Learning algorithm as the learning algorithm for the agent and design the state spaces and action spaces accordingly. To evaluate the effectiveness of our approach, we perform simulations on various MEC network topologies and compare our results with existing routing algorithms. The proposed DRL-based traffic routing achieves a balanced trade-off between overall bandwidth utilization as well as end-to-end delay, as evidenced by the convergence of the reward function to approximately 0.65 after 1000 episodes, showcasing its effectiveness in optimizing network performance Overall, the effectiveness and performance of MEC networks may be enhanced by our suggested DRL-based traffic routing method.