A Service-Based Approach To Modeling And Analyzing Trustworthy Of Cyber Physical Systems

Cyber physical system (CPS) is the system with a tight composition of computational capacities and physical processes. A set of computational units and physical entities are interlinked in a heterogeneous network. Through wide-open environment perception and deep interactive collaboration, the computation, communication and control are multidimensional intricately mixed in CPS. CPSs have much prospect in various application areas such as personal health care, intelligent building and home, intelligent transportation, smart grid, wearable device and emergency response etc. CPSs are usually deployed in safety-critical domains, and therefore they have strict requirements of real-time, distribution, dependency, fault-tolerance, security and privacy, interaction, dynamic adaptation and autonomy. The method of CPS design and development should be flexible and agile due to the complexity of the system construction and the variety of scales.We propose a service-oriented CPS architecture. Service-oriented architecture supports novel technologies like virtualization and resource allocation. And due to its loosely coupling and well-defined interfaces, the seamless integration of CPSs is applicable. In CPS, physical devices, computing units and network communications all provide certain types of services with interaction and collaboration. There may exist dependent and restraint relationship between devices because of the resource and location constraints. Therefore, the service-based CPS development is unlike the traditional service-oriented software development. From the perspective of trustworthy development, we apply the model-based development method. Through the instantiation of services, the task models of workflows can be customized according to the business processes and user requirements. Aspect-oriented method separates the concerns in complex systems. Those common general characteristics can be extracted out and form an aspect module. This helps to improve the reusability and structuralization of the system. In this thesis we apply the aspect-oriented method to the modeling of CPS on multiple concerns. With the combination of Petri net and its extension, the formal verification and analysis can be performed on the model to avoid design error and defect in the early stage of system development. Aiming at the special characteristics of CPSs, we focus on the modeling and analysis about the fusion of computation and physical processes, the time constraint and schedulability, the optimization of resource services selection and composition. Then from the perspective of runtime dependability and robustness, we propose a fault-tolerant and self-adaptation modeling and analyzing approach for CPSs. The main contributions of this thesis are summarized as follows.(1) A service-based CPS architecture is proposed which refines the three-tiered architecture model. This architecture supports top-down development and bottom-up analysis. Its capability of flexible support of business processes enables the dynamic composition and reconfiguration of the runtime system. Then a SysML/MARTE is proposed to model the requirements of CPSs. Through the well-described rules, the SysML model can be transformed into Petri nets to support the further formal verification and validation.(2) An aspect-oriented discrete and continuous Petri net is presented to model the computation and physical processes of CPSs. The common and general physical processes are extracted out as a crosscutting concern and described in an aspect Petri net. The aspect net can be weaved into the basic core concern model. Through well-defined weaving rules and input and output interfaces, the composition model can maintain its correctness and consistence.(3) A time constraint aspect-oriented Petri net (TAOPN) is proposed to model the time properties of CPSs. The time constraint is defined as a combination of time interval and time delay due to the needs of CPSs. Then a compositional schedulability analysis and algorithm is proposed. Through a case study demonstration, the approach TAOPN is applicable for the modeling of real-time system with time constraints.(4) A resource service composition model based on colored Petri net is discussed. And aspect-oriented method can be integrated into the composition for some non-functional requirements purpose. Then a resource scheduling and allocation method is discussed. Two resource types are defined and analyzed with the objective functions for the optimization. One is single objective scheduling problem which using Particle Swarm Optimization can find the best solution. The other is the multi-objectives scheduling problem. A Particle Swarm based Pareto Optimization algorithm is proposed to trade-off the multi-objectives of the resources, and the best solution of resource scheduling and allocation can be found.(5) A runtime fault-tolerant adaptation model of CPS is put forward. The model is described in two levels as devices and tasks. And the types of fault causes are discussed. Then a fault output model, the respective fault-tolerant and adaptation strategy aspect models are proposed. The aspect models can be weaved into the functional model to form fault-tolerant and adaptive models according to the fault types. The composition model is analyzed by computational tree logic method.