Performance Improvement of AODV in Wireless Networks using Reinforcement Learning Algorithms

Abstract

This paper investigates the application of reinforcement learning (RL) techniques to enhance the performance of the Ad hoc On-Demand Distance Vector (AODV) routing protocol in mobile ad hoc networks (MANETs). MANETs are self-configuring networks consisting of mobile nodes that communicate without the need for a centralized infrastructure. AODV is a widely used routing protocol in MANETs due to its reactive nature, which reduces overhead and conserves energy. This research explores three popular Reinforcement Learning algorithms: SARSA, Q-Learning and Deep Q-Network (DQN) to optimize the AODV protocol's routing decisions. The RL agents are trained to learn the optimal routing paths by interacting with the network environment, considering factors such as link quality, node mobility, and traffic load. The experiments are conducted using network simulators to evaluate the performance improvements achieved by the proposed RL-based enhancements. The results demonstrate significant enhancements in various performance metrics, including reduced end-to-end delay, increased packet delivery ratio, and improved throughput. Furthermore, the RL-based approaches exhibit adaptability to dynamic network conditions, ensuring efficient routing even in highly mobile and unpredictable MANET scenarios. This study offers valuable insights into harnessing RL techniques for improving the efficiency and reliability of routing protocols in mobile ad hoc networks.