Study On Multi-task Oriented Services Composition And Optimization In Cloud Manufacturing

The dramatic development of the state-of-the-art IT trends such as cloudcomputing, Internet of things, etc., combining with the advanced manufacturing modelsand technologies, has contributed to the emergence of cloud manufacturing. The newmanufacturing model aims to assemble and optimally allocate global distributedmanufacturing resources, which is emerging as the major enablers for thetransformation and upgrading of manufacturing industry. Services composition andoptimization (SCO) is the key technology to optimally allocate manufacturing resourceswithin cloud manufacturing system. However, as an open system, cloud manufacturingis characterized by the salient feature of multi-user manufacturing, which makes it quitecommon that the SCO is under the context of multi-task. The high flexibility of systemand complexity of requirements also lead to complexity of the multi-task requestingmodes, thereby complexing the problem of SCO in cloud manufacturing. How to adaptto the context of multi-task and the complex multi-task requesting modes, and toimprove the allocation of resource by developing and using SCO technologies, is thefocus of this dissertation.The investigation regarding the issue of SCO within cloud manufacturing will belaunched in this dissertation, with the objective of optimal allocation of manufacturingresources under the context of multi-task and the complex multi-task requesting modes,so as to find out the key technologies to improve the overall outcome of cloudcomposite services. The main content and contributions are listed as follows:①The Framework of SCO in cloud manufacturing and three kinds of compositionpatterns are proposed. Initially, The Framework of SCO is developed to catalogue thekey technologies to the problem of SCO in cloud manufacturing. Within the Framework,the related work of the key technologies is studied, whose usability is also analyzed.Then, three kinds of composition patterns are proposed after discussing the applicabilityof single-task oriented services composition and optimization (STO-SCO) in themuti-task context of cloud manufacturing, including Each Composition for Each Task(ECET) pattern, Multi-Composition for Each Task (MCET) pattern, and Multi-Composition for Multi-Task (MCMT) pattern. The transform conditions from onepattern to another are also proposed, which forms the patterns into a conceptual system.②To address the scenario of “simple multi-task requesting mode”, multi-task oriented service composition and optimization (MTO-SCO) with the ECET pattern isstudied and the related technologies are developed. The main contributions to theproblem include:1) The limitation in the number of requesting tasks in STO-SCO isremoved to make a holistic decision on multi-task composition, thereby overcoming thecomposition failure caused by ignoring the underlining QoS constraints over multi-task.2) A matrix real-coded based genetic algorithm is proposed to implement the holisticdecision-making in MTO-SCO and the simulation experiment results suggest that theproposed algorithm can solve the MTO-SCO problem both efficiently and effectively.③To address the scenario of “some requesting tasks with severe QoS constraints”,the “one-to-one” mapping between requesting tasks and cloud composite services isrelaxed, and a new problem of MTO-SCO with the MCET pattern is investigated andthe related technologies are proposed. The main contributions to the problem include:1)A progressive global approach is proposed to integrate several incompetent cloudcomposite services to collectively perform a task with severe QoS constraints; as shownin comparative experiments results, the new global approach significantly outperformsexisting global approaches on both overall QoS outcome and success rate of QoSrequirement fulfillment.2) The exterior aggregation patterns and formulas are exploitedin multi-task scenarios, which are supplements to the traditional services compositionand optimization theories.3) A hybrid-operator based matrix coded genetic algorithm(HO-MCGA) is developed to implement the three-staged holistic decision-making,which achieves a significant fitness improvement at a marginal extra time cost asopposed to the simplex-operator based matrix coded genetic algorithm.④To address the scenario of “available cloud services suffering a quantitativeshortage relative to requesting tasks”, the assumption of “a cloud composite service cannot be shared” is removed, thereby proposing a new problem of MTO-CSCO with theMCMT pattern and developing the related technologies. The investigation for theproblem of MTO-CSCO is advanced based on the following main contributions:1) amore progressive global approach is designed which allows both the integration ofincompetent cloud composite services and the optimal allocation of requesting tasks.The proposed approach outperforms the MCET pattern based global approach onoverall QoS outcome, success rate of QoS requirement fulfillment, and utilization rateof manufacturing resources, especially as the available cloud services are insufficient.2)A hybrid-operator based collaborative evolutionary matrix coded genetic algorithm(HOCE-MCGA) is developed, which surpasses HO-MCGA in both effectiveness and efficiency.The aforementioned investigations and contributions forms a methodologyembracing composition patterns, global approaches, QoS modeling technologies andintelligent algorithms, for investigating MTO-CSCO problem in the context of cloudmanufacturing.