RLbRR: A Reliable Routing Algorithm Based on Reinforcement Learning for Self-organizing Network

Abstract

AbstractWith wireless communication rapid development, self-organizing network has been applied widely in various fields. Self-organizing network routing algorithm is a significant research direction. This work explores a reinforcement learning approach applied to self-organizing network packet routing. It is found that the existing routing algorithms for self-organizing network always use a single metric for next hop selection and fail to consider the influence of neighboring nodes on the forwarding nodes. The selection of the next hop does not have a comprehensive and long-term consideration, which is detrimental to promote routing reliability. Reinforcement learning can be used to solve this problem. In this paper, a Reinforcement Learning based Reliable Routing (RLbRR) algorithm is designed after studying to the existing methods in the literatures. RLbRR algorithm evaluates the quality of the forwarding node by reinforcement learning algorithm and combines with multi components. In this way, the selection of the next hop can become all-sided, so as to establish a stable and reliable routing path. The simulation results show that RLbRR algorithm performs well in reliability.KeywordsRouting algorithmReinforcement learningSelf-organizing network