中文摘要：

首先搭建认知无线电网络环境，在该网络下分析TCPReno、TCPNewreno、TCPSackl、TCPVegas、TCPWestwood协议的性能包括拥塞窗口、平均吞吐量、丢包率。针对传统TCP拥塞控制在认知无线网络中存在的问题，提出一种跨层优化TCPWestwood的改进算法，称为TCP-CR。该算法区分网络拥塞导致的超时和切换所导致的超时，同时针对不同轻度拥塞进行不同的恢复策略。ns-仿真结果表明，TCP-CR可以减少TCP的平均慢启动次数，吞吐量比传统的TCPWestwood显著提高。增强了对认知无线网络环境的适应性，从而提高了网络的性能。

英文摘要：

Firstly, the cognitive radio network environment is built. Subsequently, the performance of protocol TCP Reno, TCP Newreno, TCP Sackl, TCP Vegas, TCP Westwood are analyzed including Congestion window, average throughput and packet loss rate. Finally, according to the problems of traditional TCP congestion control in cognitive wireless network, a improved a- rithmetic of cross layer optimized the TCP Westwood is proposed, which called TCP-CR. This algorithm distinguishs the over- time of network congestion and switching, and uses the different recovery strategy according to different mild congestion. Using ns-2 system simulation, the result demonstrates TCP-CR can reduce the times of the TCP average slow starts, the throughput rate can increase significantly than the traditional TCP Westwood. The algorithm strengthen the adaptability of cognitive wireless network environment and improve the network performance.