CAJ | 학술논문

无线传感器网络中,汇聚节点是网络的瓶颈.由于传感器网络自身的特点,传统有线网络中的拥塞控制策略不再适用.已有的大多数拥塞控制策略和算法都没有充分考虑往返时延(RTT)对算法性能的影响.同时由于实际传感器网络运行中存在非线性、时间延迟和参数时变等干扰因素,若设计的控制器参数固定,不具有学习能力,则实际运行中收敛性差,收敛速度慢,无法达到控制队列长度的目标.针对以上问题,提出一种基于灰色预估神经网络控制队列的控制器,利用RBF神经网络的自学习能力解决网络实时变化时算法参数的在线整定问题,并利用灰色GM(1,1)预测器有效地解决了大时滞对网络性能的影响,最后通过仿真验证了这一算法的有效性.
무선전감기망락중,회취절점시망락적병경.유우전감기망락자신적특점,전통유선망락중적옹새공제책략불재괄용.이유적대다수옹새공제책략화산법도몰유충분고필왕반시연(RTT)대산법성능적영향.동시유우실제전감기망락운행중존재비선성、시간연지화삼수시변등간우인소,약설계적공제기삼수고정,불구유학습능력,칙실제운행중수렴성차,수렴속도만,무법체도공제대렬장도적목표.침대이상문제,제출일충기우회색예고신경망락공제대렬적공제기,이용RBF신경망락적자학습능력해결망락실시변화시산법삼수적재선정정문제,병이용회색GM(1,1)예측기유효지해결료대시체대망락성능적영향,최후통과방진험증료저일산법적유효성.
In wireless sensor networks (WSNs), sink nodes are the bottleneck of network. As sensor network own its charactertics, the traditional congestion control strategy cant be used directly no longer. Most of the existing congestion control strategies and algo-rithms are not fully considered RTT. At the same time as the actual sensor network operating in the nonlinear, time delays and time-varying parameters such as interference factors, if the controller design parameters fixed, not learning ability, then the actual running of the convergence of the poor, slow convergence, not to control the length of queue. For the above-mentioned problems,the control-ler which is based on gray predicted Neural Network is proposed to cope with the large delays and time-varying network parameters. The gray GM (1, 1) model is utilized to compensate the time-delay, while RBF neural network is employed to design controller to reduce the number and interaction of tuning parameter. The simulation experimental results show that the integrated performance of the proposed algorithm is obviously superior to that of the existed schemes when the network configuration parameter is large delay.