Optimizing the Routing in Ad Hoc Networks With Option-Critic

The mobile Ad Hoc networks require routing algorithms with fast convergence speed, strong dynamic adaptability, and high information fault tolerance due to the characteristics of strong node mobility and dynamic change of topology. However, most of the existing routing algorithms fail to solve the problems caused by the dynamically changing characteristics of mobile Ad Hoc networks. To deal with this problem, this paper proposes an intelligent routing algorithm which can dynamically optimize network routing. Especially, we adopt option-critic architecture based on hierarchical reinforcement learning to train agents. The proposed algorithm can optimize routing with few a priori conditions. Simulation results show that the proposed algorithm has better convergence and effectiveness than the existing Q-learning algorithm, and provides better routing configuration, thereby improving network performance.