一种认知无线网络中TCP拥塞窗口的在线学习方法

Online Learning Method for TCP Congestion Control in Cognitive Radio Networks

在认知无线电中,改进传输层协议提高吞吐量的方法引起了业界的广泛关注。采用可用带宽测量技术,提出了一种基于ACK时间间隔的在线学习方法(TCP-Learning)。该方法能够快速地学习到网络链路中可用剩余带宽;并能够快速调整TCP拥塞窗口。仿真结果表明,在链路处于良好状态下,TCP-Learning吞吐量略优于Reno和Vegas等传统拥塞控制算法,但在链路较差情况下,TCP-Learning吞吐量明显优于Reno和Vegas等传统拥塞控制算法。

Improved transport layer protocol have been widespread concerned to improve the throughput in cog- nitive radio networks. An online method （TCP-Learning） based on ACK interval is proposed which is used to probe the available bandwidth of TCP. The method can learn available residual bandwidth in the network and quickly ad- just the congestion window of TCP. The simulation results show that the throughput of TCP-Learning performs the traditional congestion control algorithm such as Reno and Vegas in bad link conditions.