无线网络中基于强化学习的拥塞控制算法改进  

Improvement of the Congestion Control Algorithm Based on Reinforcement Learning in Wireless Network

作　　者：罗颖[1] 林茂松[1] 江虹[1] 

机构地区：[1]西南科技大学信息工程学院,四川绵阳621010

出　　处：《自动化仪表》2014年第6期65-68,共4页Process Automation Instrumentation

基　　金：国家自然科学基金资助项目(编号:61072138);研究生创新基金资助项目(编号:14ycxjj0102)

摘　　要：针对现有协议在无线网络中出现的吞吐量小、传播延迟大等问题,对TCP-NewReno协议进行研究,提出了一种改进的拥塞控制协议(NewReno-RF)。考虑无线网络拓扑变化快、带宽不对称的特性,在拥塞控制协议的慢启动和拥塞避免阶段,提出了基于往返时延量化和强化学习的拥塞窗口自适应动态变化算法,以对拥塞窗口进行速率控制。NewReno-RF算法在NS2中建立的无线拓扑网络仿真结果表明,其较TCP-NewReno明显改善了无线网络的通信质量。The existing protocols features disadvantages in wireless networks, e. g. , less throughput and large time delay in propagation, aiming at this situation, the TCP-NewReno protocol has been researched, and the improved congestion control protocol NewReno-RF is proposed. Considering the rapid changes of the topology and asymmetric bandwidth of the wireless network, the congestion window （ cwnd ） adaptive dynamic variation algorithm based on back and forth delay quantization and reinforcement learning is proposed to control changing rate of congestion window in slow start and congestion avoidance stages of the congestion control protocol. Wireless topological network is established with NewReno-RF in NS2, the result of simulation shows that comparing with TCP-NewReno, NewReno-RF obviously improves the communication quality of wireless network.

关 键 词：TCP AIMD拥塞控制 NewReno—RF 量化 强化学习 Q学习 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]