中文摘要：

<正>Active queue management(AQM) is essentially a router buffer management strategy supporting TCP congestion control.Since existing AQM schemes exhibit poor performance and even instability in time delay uncertain networks,a robust buffer management(RBM) mechanism is proposed to guarantee the quality of service(QoS).RBM consists of a Smith predictor and two independent controllers.The Smith predictor is used to compensate for the round trip time(RTT) delay and to restrain its negative influence on network performance.The main feedback controller and the disturbance rejection controller are designed as proportional-integral (PI) controller and proportional(P) controller by internal model control(IMC) and frequency-domain analysis respectively.By simulation experiments in Netwrok-Simulator-2(NS2),it is demonstrated that RBM can effectively manage the buffer occupation around the target value against time delay and system disturbance. Compared with delay compensation-AQM algorithm(DC-AQM),proportional-integral-derivative(PID) algorithm and random exponential marking(REM) algorithm,the RBM scheme exhibits the superiority in terms of stability, responsiveness and robustness.

英文摘要：

Active queue management （AQM） is essentially a router buffer management strategy supporting TCP congestion control. Since existing AQM schemes exhibit poor performance and even instability in time delay uncertain networks, a robust buffer management （RBM） mechanism is proposed to guarantee the quality of service （QoS）. RBM consists of a Smith predictor and two independent controllers. The Smith predictor is used to compensate for the round trip time （RTT） delay and to restrain its negative influence on network performance. The main feedback controller and the disturbance rejection controller are designed as proportionalintegral （PI） controller and proportional （P） controller by internal model control （IMC） and frequency-domain analysis respectively. By simulation experiments in Netwrok-Simulator-2 （NS2）, it is demonstrated that RBM can effectively manage the buffer occupation around the target value against time delay and system disturbance. Compared with delay compensation-AQM algorithm （DC-AQM）, proportional-integral-derivative （PID） algorithm and random exponential marking （REM） algorithm, the RBM scheme exhibits the superiority in terms of stability, responsiveness and robustness.