Research Of QoS Routing Algorithm Based On Reinforecment Learning In Moblie Ad Hoc Network

A Mobile Ad Hoc Network(MANET) is an autonomous multi-hop system of Mobile nodes connected by wireless links without the support of static infrastructure such as base station. The demanding for supporting different quality of service (QoS) in the network arises. Especially the audio, video and real time data communication applications require strict QoS guarantee such as delay, delay jitter, priority differentiation and Bandwidth Allocation.In the dissertation, based on summarizing previous research, a systematical and in depth study has been done in solving the problems of QoS routing in Ad Hoc networks by using reinforcement learning algorithm:1. Existing mobile ad-hoc routing protocols are based on a discrete, bimodal model for links between nodes: a link either exists or is broken. This model cannot distinguish fail transmissions which fail due to interference or congestion from those due to their target being out of transmission range. A statistical network link model is introduced to represent the quality of the link by a statistical measure of link performance. A new routing protocol called SNLQ is proposed based on a combination of continuous (rather than discrete) model for links and the reinforcement learning within the Ad Hoc network. Different scenario-based performance evaluations of the protocol in NS-2 are presented. In comparisons with AODV and DSR, SNLQ routing exhibits improved performance in congested wireless networks.2. Studying the math model of multi QoS constraint routing and using reinforcement learning algorithm, this dissertation points out an algorithm named SARSA-R, which uses reinforcement learning algorithm to solve multi QoS constraint problems In Ad Hoc. This algorithm considers some QoS metrics such as node battery energy,Link bandwidth,delay and packet lost rate. Simulation experiments prove that this algorithm plays good capability on multi-QoS constraint problems.3. Using NS2 network simulator as the simulation platform, I designed a simulation system combine with SNLQ and SARSA-R. The simulation results show that the algorithm proposed in this paper take full advantage of limited bandwidth of Ad Hoc networks , reduce the average end-to-end transmission delay, and the objective and subjective video quality of the received video have markedly improved.