Dynamic Routing Path Selection Algorithm Using Reinforcement Learning in Wireless Ad-Hoc Networks

초록

This paper proposes a routing protocol scheme using reinforcement learning which supports dynamic wireless communication conditions in ad-hoc networks. The aim of this scheme is to maximize the utility value of routing path in terms of transmission rate, residual energy and end-to-end delay. Q-learning based routing path selecting algorithm is proposed with consideration of packet successful transmission ratio. Each node represents a state and the next packet transmission path link between nodes is called an action in Q-learning. A reward is given to visited path when a packet reaches to destination based on reliability of packet transmission and energy consumption. The simulation results show that our method can obtain dynamic environmental adaptivity and high utility in various communication situations.

키워드