太赫兹科学与电子信息学报
太赫玆科學與電子信息學報
태혁자과학여전자신식학보
Information and Electronic Engineering
2013年
6期
970-976,980
,共8页
Markov决策过程%Markov切换系统%性能势%策略迭代
Markov決策過程%Markov切換繫統%性能勢%策略迭代
Markov결책과정%Markov절환계통%성능세%책략질대
Markov decision process%Markov switching system%performance potentials%policy iteration
基于3层架构的映射框架，本文以视频业务为主要研究对象，研究了分布式协同接入控制系统的建模和调控机理。不失一般性，假设视频业务需求近似服从泊松分布，而服务时间服从指数分布，将一个资源节点上的业务接入状态的演化用 Markov过程来描述，则所有的 N个资源节点的分布式协同接入控制可以由一个 Markov 切换空间控制过程来阐述。针对构建的 Markov切换空间控制系统，基于系统灵敏度的观点对系统性能进行优化，本文提出了基于策略梯度的随机逼近算法；同时，证明了该随机逼近算法的收敛性。此外，本文给出了一个仿真实例来说明此随机逼近算法的有效性。
基于3層架構的映射框架，本文以視頻業務為主要研究對象，研究瞭分佈式協同接入控製繫統的建模和調控機理。不失一般性，假設視頻業務需求近似服從泊鬆分佈，而服務時間服從指數分佈，將一箇資源節點上的業務接入狀態的縯化用 Markov過程來描述，則所有的 N箇資源節點的分佈式協同接入控製可以由一箇 Markov 切換空間控製過程來闡述。針對構建的 Markov切換空間控製繫統，基于繫統靈敏度的觀點對繫統性能進行優化，本文提齣瞭基于策略梯度的隨機逼近算法；同時，證明瞭該隨機逼近算法的收斂性。此外，本文給齣瞭一箇倣真實例來說明此隨機逼近算法的有效性。
기우3층가구적영사광가，본문이시빈업무위주요연구대상，연구료분포식협동접입공제계통적건모화조공궤리。불실일반성，가설시빈업무수구근사복종박송분포，이복무시간복종지수분포，장일개자원절점상적업무접입상태적연화용 Markov과정래묘술，칙소유적 N개자원절점적분포식협동접입공제가이유일개 Markov 절환공간공제과정래천술。침대구건적 Markov절환공간공제계통，기우계통령민도적관점대계통성능진행우화，본문제출료기우책략제도적수궤핍근산법；동시，증명료해수궤핍근산법적수렴성。차외，본문급출료일개방진실례래설명차수궤핍근산법적유효성。
The modeling of distributed collaborative admission control system and mechanism are studied based on the three-layer structure of mapping framework. The video service is considered as the main research object, which is assumed approximating Poisson distribution, and service time obeys exponential distribution. A resource node in the evolution of the state of the service can be described in Markov process. Accordingly, all the N resource nodes of distributed collaborative admission control can be elaborated by a Markov switching space control process. According to the proposed Markov switching system, the performance of the system is optimized from the viewpoint of sensitivity. A stochastic approximation algorithm is put forward based on the strategy of the gradient. Meanwhile, the convergence of the algorithm is proved, and an example is given to verify the effectiveness of the algorithm.
