Research On Key Technologies Of CPS-Service Run-time Supporting Platforms

Cyber physical systems(CPSs)are a kind of multidimensional and complicated systems that incorporate computation,communication and physical subsystems.In the today's IOE(Internet of Everything)age,CPSs have broad application prospects.However,the inherent complexity of CPSs challenges their design and implementation.First,there are many physical entities(PEs)of a great variety and high heterogeneity in CPSs,thus it is very difficult to manage them by using a unified way.Second,CPS tasks cannot be determined in advance,so it is necessary to provide run-time environments to support the execution of such dynamic tasks.Third,the emergence of the unstructured and unknown environments demands the high intelligence of PEs,however current PEs always do not have powerful computational capability to enable it.As a sequence,developing CPS applications from scratch is inefficient.We,according to the features and requirements of multi-robot cooperation systems,propose to design a CPS-Service run-time supporting platform to increase the development efficiency of such CPS application systems.The platform uses CPS-Services to manage all the PEs in CPSs,and employ the CPS-Service composition technology to facilitate the execution of CPS tasks,and exploit the cloud computing technology to increase the computational capability of PEs and thus enhance CPS-Services.Several research works on the architecture and key technologies of the CPS-Service runtime supporting platform are conducted in this dissertation.These research works and the contributions of this dissertation are summarized as follows.First,we research the design of the CPS-Service run-time supporting platform.With the deep analysis about the popular architectures and related technologies of CPS run-time supporting platforms,we compare existing CPS run-time supporting platforms from two aspects:construction approaches for CPS tasks and applicable scenarios.On the basis of these,we design a real-time cloud technology based CPS-Service run-time supporting platform.Second,we propose an ontology based CPS-Service model and a corresponding CPSService matchmaking algorithm.Because current service models cannot express the particular properties of CPS-Services,we propose an ontology based CPS-Service model to deal with it.Compared to current service models,the new model is more complete,which includes several new concepts,such as physical entity,physical location,and physical environment.In addition,there are many CPS-Services managed by the CPS-Service run-time supporting platform,but the execution of a CPS task only needs a part of the CPS-Services,so to quickly find the required CPS-Services becomes to be a problem.To deal with it,we also propose a probabilistic clustering and R-tree based CPS-service matchmaking algorithm.The algorithm uses a hybrid index structure to speedup the online CPS-Service matchmaking process.Through simulations,we demonstrate that the algorithm greatly increases the speed of CPS-Service matchmaking.Third,we propose physical location and environment sensitive CPS-Service composition methods.Because current software service composition algorithms cannot handle the particular characteristics of CPS-Services,such as the physical location and environment sensitivity,we research two typical CPS-Service composition problems.One is the CPS-Service composition problem considering the physical location sensitivity;the other is the CPS-Service composition problem considering the execution reliability.A heuristic approach based on an improved quantum genetic algorithm and a hyperheuristic approach based on Annealing Pareto Thompson sampling are proposed to solve them,respectively.The simulation results show that the two algorithms outperform baseline methods under most situations.Fourth,we propose two cloud based CPS-Service enhancing methods.To handle the unstructured and unknown environments,CPS-Services need to incorporate computation-intensive tasks to enhance themselves,however current PEs always have limited computational capability.To deal with it,we propose to introduce the cloud computing technology to increase the computational capability of PEs and thus enhance CPS-Services.Nonetheless,obtaining optimal CPS-Service enhancing schemes is a nontrivial task.Two CPS-Service enhancing algorithms are presented to efficiently solve the problem.Through numerical simulations,we demonstrate that the proposed algorithms both can give quite good CPS-Service enhancing schemes,and the greedy selection based CPS-Service enhancing algorithm even can give nearoptimal CPS-Service enhancing schemes.Fifth,we implement the main components of the CPS-Service run-time supporting platform prototype system,such as,the CPS-Service enhancing framework,the CPS-Service matchmaking framework,the CPS task execution engine,and also develop two related software tools,i.e.the CPS-Service description tool,the CPS task construction tool.