基于二次型优化的神经元PID无线拥塞算法研究

PID congestion control based on neural network in Ad Hoc Network

神经元PID算法能较好地控制瓶颈节点的队列长度,但当网络环境发生较大变化时,其控制效果往往难以保证.根据Ad Hoc网络环境参量时变的特点,推导了无线TCP/AQM离散模型,在神经元算法的加权系数中引入二次型性能指标,设计了一种基于二次型性能指标的神经元PID的AQM.仿真结果表明:在动态拓扑、突发流及链路容量变化的Ad Hoc网络中,该改进算法优于PI算法.

Neuron PID can regulate the queue size of a bottleneck node successfully, but its control often is worse when the network state changes. For Ad Hoc network＇s time-varying, it deduced discrete math model of TCP/AQM control object in wireless network. It introduced Quadratic Performance Index to weighting coefficient modification for neuron PID controller. Finally, a neuron Proportional-Intions demonstrate that new algorithm performance is better than PI under different link capacity, dy-amical topology and sudden flow condition.