Deep Reinforcement Learning Routing in Mobile Networks

Abstract

In modern computer networks, billions of devices are connected to the Internet. Traditional routing algorithms have not been designed for modern networks. In modern computer networks, traditional routing algorithms fail to combat congestion dynamically, and they cannot satisfy modern Quality of Service (QoS) needs. To overcome this problem, a deep reinforcement learning-based routing method has been proposed. Routing decisions rely on neural network architectures, instead of traditional heuristic algorithms. Our proposed method was compared to the Dijkstra shortest-path algorithm in a mobile network, and our method outperformed the Dijkstra algorithm. © 2025 Elsevier B.V., All rights reserved.