Research of QoS Routing Algorithm in Ad Hoc Networks based on Reinforcement Learning

摘要

With the prevalence of multimedia application, it has become a research focus to provide QoS in ad hoc mobile network. According to the features of recent routing algorithms over ad hoc network, such as the discrete, bimodal model for links between nodes, a new routing algorithm called SNLQ is proposed based on continuous link model with reinforcement learning literature. More concretely, it moves the method of calculating Q-values onto link-values with an eye to a combination of the fixed-time retransmissions mechanism in 802.11MAC and the continuous state-values and Q-values in reinforcement learning. Different scenario-based performance evaluations of the protocol in NS-2 are presented. The results show that our algorithm effectively improves the link table and considerably increases the packet delivery ratio which is superior to AODV and DSR in the congested wireless networks.