基于遗传预估策略的智能主动队列管理

针对TCP传输过程中的典型时滞特性,提出了一种智能主动队列管理算法.该算法以自学习预估机制模型为核心来克服大时滞特征对网络稳定性能的影响,拥塞控制系统以两条信息通道分别实现模型补偿和预测控制功能.模型补偿通道采用了Smith预估嚣实现对网络时滞特征的动态补偿,并进一步设计迭代进化算法实现对Smith预估模型未建模特征的估计过程.预测控制通道采用基于神经网络的PID智能丢弃算法,通过神经网络的学习预测功能自适应调整预测控制通道的控制行为.通过仿真研究表明了提出的控制方法显著提高了拥塞控制机制的稳定性能和自适应性能.

Genetic-predictor based intelligent active queue management

Aiming at the time-delay characteristic presented in the TCP transmission process,an intelligent active queue management(AQM) scheme is proposed to handle the large time-delay features with a self-learning predictor model.The congestive control includes two information channels for the model compensation and predictive control of AQM.The model compensation channel uses a Smith predictor to achieve the dynamical compensation for the large time-delay,and the iterative genetic algorithm is used to estimate the un-modeled parameters.The predictive control channel uses a neural network based PID discarding algorithm,and the self-learning characteristic of the neural network is integrated in the design of PID.The control behavior is auto-tuned by the adaptive neural network.Simulation results show that the proposed method can improve the stability and adaptive performance greatly.