CPS资源服务模型和资源调度研究

Research on CPS Resource Service Model and Resource Scheduling

信息物理融合系统(Cyber-Physical Systems,CPS)融合了信息世界与物理世界,作为工业互联网与智能制造等重要领域的关键技术引起了越来越多的关注.与WEB服务相比,CPS系统中资源种类繁多、数量庞大、资源之间的异构性强,并且具有大量重复的物理实体,不同物理实体所处的物理环境不同且具有不同的执行性能(QoS),导致在对具体任务进行资源调用时存在多种调度方案,该文针对该问题展开研究.首先,鉴于CPS系统结构复杂的特点,该文综合分析现有的CPS建模研究成果,采用面向服务的体系架构(Service-Oriented Architecture,SOA)的思想研究CPS的体系结构,在此基础上研究了CPS与WEB服务之间的异同点,确定了面向服务CPS资源服务模型研究的可行性,提出了CPS系统中资源、物理实体、虚拟物理实体、虚拟资源的划分.同时,基于此划分提出了一种OWL(Web Ontology Language)和XML(Extensible Markup Language)混合式的CPS资源服务模型,该模型采用OWL分别对虚拟物理实体、物理实体提供的服务、CPS上层任务进行描述,并采用XML对真实物理实体的QoS参数进行描述.其次,在CPS资源服务模型的基础上,提出了基于智能规划的CPS任务-虚拟资源调度机制,该机制主要确定满足任务需要的多个候选资源集,该候选资源集组成资源的初始调度序列.然后,该文根据资源的QoS要求建立了多目标线性规划的资源选择数学模型并对多目标遗传算法的关键步骤进行创新,提出了以基于多目标遗传算法的资源选择算法求解该模型的方法,所得最优解即为资源候选集中满足整体QoS最优的资源调度序列.最后,以智能电网为例,验证了该文提出的基于智能规划的CPS任务-虚拟资源调度机制和基于多目标遗传算法的CPS资源选择方法的有效性.

Cyber-Physical Systems(CPS)combines the cyber space and the physical world.As the key technology of some important areas like Industrial-Internet and Intelligent-Manufacturing,it has attracted more and more attentions.There are more varieties of resources among which strong heterogeneities are introduced in CPS system than in WEB system.Due to the existence of a large number of duplicate physical entities which have different physical environments and different Quality of Service(QoS)respectively,multiple selectable schemes exist in the scheduling of task resources and this thesis deals with this problem.Firstly,in term of the complex structures of CPS system,this thesis comprehensively analyzes the existing research results of CPS modeling and adopts the idea of Service-Oriented Architecture(SOA)to study the CPS architecture.On the basis,this thesis also studies the similarities and differences between CPS and WEB service,confirms the research feasibility of service-oriented CPS resource service model and proposes the divisions of resources,physical entities,virtual physical entities and virtual resources in CPS.Resources are representations in cyber space of the service provided by entities in CPS.Physical entities correspond to the concrete resources in CPS.Multiple physical entities of the same type which have the same or similar functions,the different physical environment information and different QoS parameters can be described by virtual physical entities in order to reduce the number of physical entities in the model for faster search speed.Virtual resources are equivalent to virtual physical entities and a virtual resource corresponds to one resource or more in cyber space.Meanwhile,based on the divisions of resources,physical entities,virtual physical entities and virtual resources in CPS,a hybrid CPS resource service model of OWL and XML was proposed.This model uses OWL to describe the services provided by virtual physical entities,physical entities,and CPS upper-layer tasks respectively.And it uses XML to describe the QoS parameters of real physical entities.The model description method proposed by this thesis combines descriptions of the cyber space and the physical world,links up the both,connects the discrete part in cyber space and the continuous part in physical space,reflecting the logic of the cyber space and the reality of the physical world.Secondly,on the basis of CPS resource service model,a CPS task-virtual resource scheduling mechanism based on intelligent planning was proposed.This mechanism identifies multiple candidate resource sets that meet the needs of the task which comprise the initial scheduling sequence of resources.Thirdly,according to the QoS requirements,this thesis builds the resource selection mathematic model based on multi-objective linear programming and makes improvements on the critical steps of multi-objective genetic algorithm.And a method to solve the model was proposed,which is based on multi-objective genetic algorithm for resource selection algorithm.The optimal solution obtained through this method is the resource scheduling sequence that meets the overall QoS optimization in candidate resource sets.Finally,take smart grid for example,the effectiveness of CPS task-virtual resource scheduling mechanism based on intelligent planning and CPS resource selection based on multi-objective genetic algorithm was verified.