The bursty events of available bit rate (ABR) traffic challenge the research on flow control algorithm. This bursty nature makes the network difficult to guarantee some performance such as decreasing cell loss especia...The bursty events of available bit rate (ABR) traffic challenge the research on flow control algorithm. This bursty nature makes the network difficult to guarantee some performance such as decreasing cell loss especially when congestion appears. In addition the robust stability analysis in the presence of virtual connections (VCs) variation is another inherent problem of the current rate-based algorithm. An explicit rate-based scheme is concerned. Some mathematical methods are adopted to minimize overshoot in the buffer to eliminate the growing congestion quickly. Then in linear matrix inequality (LMI) forms via Lyapunov stability argument relating to robustness issues in the presence of time-varying VCs is purposed in linear matrix inequality. A set of numerical examples guarantees this choice of algorithm parameters is robust stability. The simulations further back up our results.展开更多
为了在802.11的网络中提供服务质量(QoS)支持,IEEE 802.11 Task Group E提出了EDCF协议.然而EDCF只能提供业务区分服务,并不能提供服务质量(QoS)保证.为了能在重负载下提供QoS保证,在WLAN中加入呼叫接纳控制(CAC)机制是非常必要的.本文...为了在802.11的网络中提供服务质量(QoS)支持,IEEE 802.11 Task Group E提出了EDCF协议.然而EDCF只能提供业务区分服务,并不能提供服务质量(QoS)保证.为了能在重负载下提供QoS保证,在WLAN中加入呼叫接纳控制(CAC)机制是非常必要的.本文首先提出了一个新的3维Markov模型对非饱和状态下EDCF的吞吐量和平均接入时延进行了分析.并在此基础上,提出了一种基于效用函数的CAC策略,它可以使网络的总收益达到最大.最后通过大量仿真验证了所提出的CAC策略的有效性.展开更多
研究多站点传送带给料生产加工站(Conveyor-serviced production station,CSPS)系统的最优控制问题,其优化目标是通过合理选择每个CSPS的Look-ahead控制策略,实现整个系统的工件处理率最大.本文首先根据多Agent系统的反应扩散思想,对每...研究多站点传送带给料生产加工站(Conveyor-serviced production station,CSPS)系统的最优控制问题,其优化目标是通过合理选择每个CSPS的Look-ahead控制策略,实现整个系统的工件处理率最大.本文首先根据多Agent系统的反应扩散思想,对每个Agent的原始性能函数进行改进,引入了具有扩散功能的局域信息交互项(原始项看作具有反应功能);并运用性能势理论,构建一种适用于平均和折扣两种性能准则的Wolf-PHC多Agent学习算法,以求解决策时刻不同步的多站点的协作Look-ahead控制策略.最后,论文通过仿真实验验证了该算法的有效性,学习结果表明,通过性能函数的改进,各工作站的负载平衡性得到改善,整个系统的工件处理率也明显提高.展开更多
文摘The bursty events of available bit rate (ABR) traffic challenge the research on flow control algorithm. This bursty nature makes the network difficult to guarantee some performance such as decreasing cell loss especially when congestion appears. In addition the robust stability analysis in the presence of virtual connections (VCs) variation is another inherent problem of the current rate-based algorithm. An explicit rate-based scheme is concerned. Some mathematical methods are adopted to minimize overshoot in the buffer to eliminate the growing congestion quickly. Then in linear matrix inequality (LMI) forms via Lyapunov stability argument relating to robustness issues in the presence of time-varying VCs is purposed in linear matrix inequality. A set of numerical examples guarantees this choice of algorithm parameters is robust stability. The simulations further back up our results.
文摘为了在802.11的网络中提供服务质量(QoS)支持,IEEE 802.11 Task Group E提出了EDCF协议.然而EDCF只能提供业务区分服务,并不能提供服务质量(QoS)保证.为了能在重负载下提供QoS保证,在WLAN中加入呼叫接纳控制(CAC)机制是非常必要的.本文首先提出了一个新的3维Markov模型对非饱和状态下EDCF的吞吐量和平均接入时延进行了分析.并在此基础上,提出了一种基于效用函数的CAC策略,它可以使网络的总收益达到最大.最后通过大量仿真验证了所提出的CAC策略的有效性.
文摘研究多站点传送带给料生产加工站(Conveyor-serviced production station,CSPS)系统的最优控制问题,其优化目标是通过合理选择每个CSPS的Look-ahead控制策略,实现整个系统的工件处理率最大.本文首先根据多Agent系统的反应扩散思想,对每个Agent的原始性能函数进行改进,引入了具有扩散功能的局域信息交互项(原始项看作具有反应功能);并运用性能势理论,构建一种适用于平均和折扣两种性能准则的Wolf-PHC多Agent学习算法,以求解决策时刻不同步的多站点的协作Look-ahead控制策略.最后,论文通过仿真实验验证了该算法的有效性,学习结果表明,通过性能函数的改进,各工作站的负载平衡性得到改善,整个系统的工件处理率也明显提高.
