动态优化是计算机系统与计算机网络中进行资源分配与任务调度等方面研究所采用的主要理论工具之一.目前,国内外已开展大量研究,致力于深化动态优化的理论研究与工程应用.文中从模型、求解与应用3个角度,对马尔可夫决策过程动态优化理论模型进行了综述,并重点介绍了将动态优化理论与随机Petri网理论相结合的马尔可夫决策Petri网和随机博弈网模型,详细讨论了这些模型的建模方法、求解算法与一些应用实例.最后,对全文进行了总结,并对未来可能的研究方向进行了展望.
Dynamic optimization is one of the most popular theoretical tools to study resource allocation and task scheduling problems in computer systems and computer networks.At present,a vast number of researches have been on their way to enhance the theoretical basis and extend the industrial applications of dynamic optimization theory.This paper provides an overview of Markov Decision Process(MDP) from the perspectives of models,solutions,and applications.We also survey two types of extended dynamic optimization models,i.e.,Markov Decision Petri Nets(MDPN) and Stochastic Game Nets(SGN),which combine dynamic optimization theory and stochastic Petri nets theory.We focus on the model construction,solution techniques,and applications of these models.Finally,we discuss some possible research challenges in the future.