【摘要】 TCP/AQM系统是一多时滞回路耦合时变系统,源端传输控制协议(TCP)的拥塞控制机制也各不相同,因此其机理模型复杂.设计控制器所采用动态近似模型的参数变化范围很大,这对辨识算法提出了很高要求.本文先从模型误差分配方式这一新的角度对经典辨识算法进行了分析.然后运用新分析框架,结合TCP/AQM系统特征和控制要求,设计了一种基于模型有效性评价机制的辨识算法.该辨识算法能够对控制系数进行自适应调节,满足了主动队列管理(AQM)算法的设计需求.仿真结果表明,基于模型有效性评价机制的AQM算法,能有效应对时滞和不确定性带来的负面影响.与潜在标准化AQM算法相比,新算法在保持低排队时延和高效链路利用率方面取得了更好的平衡.更多还原

【Abstract】 TCP/AQM system is a multi-delay coupled system with different endpoint congestion control mechanisms,hence it is difficult to accurately model this plant. Parameters of the dynamic approximate model adopted to design controllers vary widely, which puts high pressure on the identification algorithm. This paper first analyses some classical identification algorithms from a new perspective of model error allocation. Based on the novel analytical framework and considering TCP/AQM system’s characteristics and control requirements, a new identification algorithm based on model effectiveness evaluation is designed. The proposed algorithm adjusts the control coefficient adaptively, meeting the design requirements of active queue management(AQM) algorithm. Simulation results show the AQM algorithm based on the model effectiveness evaluation can deal with the negative effects brought by delay and uncertainty. Compared with the potential standardized AQM algorithms, the new algorithm achieves a better balance between maintaining low queueing delay and efficient link utilization.更多还原