基于DEC_POMDP的Web服务组合优化算法

摘要

面向服务计算( SOC)和面向服务架构( SOA)技术共同推动了Web服务及其组合技术的发展。网络环境的动态变化及其对Web服务质量( QoS)的影响，给服务成功组合带来挑战，为服务组合效果满足用户需求带来难题。为了得到经济、省时且成功率高的服务组合策略，综合考虑网络环境的动态变化、服务质量的可变性、用户需求的多样性，采用分散的部分可观测马尔可夫决策( DEC_POMDP)模型描述多个服务Agent的自组织服务组合系统，在基本Q学习算法基础上做出改进，求解模型得到组合策略。实验结果表明求解的策略较大地提高了组合服务的成本、时间消耗，且组合成功率较高。采用的DEC_POMDP模型有效地将Web服务组合动态过程描述出来，并自适应地更新了QoS值，采用Q学习算法及时使用了最新的QoS值。%Service-Oriented Computing ( SOC) and Service-Oriented Architecture ( SOA) promoted Web service and its composition of technology development. Changes in the network environment and its impact on Quality of Service ( QoS) ,brought challenge to the suc-cessful combination and problems to the service composition effect to meet customer needs. In order to get economic,time-saving and high success-rate service composition strategy,considering the changes in the network environment,variability of QoS and the diversity of user needs,use DEC_POMDP model to describe the service Agent combination of self-organizing systems,making an improvement based on Q learning algorithm to solve composition of the model. Experimental results show that solving strategy largely improves the services composition cost,time-consuming and the composition success rate is higher. The DEC_POMDP model describes the dynamic process of Web service composition and adaptively updates the QoS values,the use of Q-learning algorithm makes it in a timely manner using the latest QoS values.