Research On Manufacturing Cloud Services Composition Based On Extended HTN

In the influence of Advanced Manufacturing Technology(AMT), NetworkManufacturing, Manufacturing Grid et al manufacturing schema, the notion of CloudManufacturing has emerged with the development of information technology and theInternet of Things. In the environment of Cloud Manufacturing, everything is serviceand it needs manufacturing cloud service composition which users can usemanufacturing resources and capability on demand. On the other hand, traditionalservices composition can not meet dynamic and efficient demand of cloudmanufacturing because of complex environment and the characters of manufacturingindustry. Therefore, researching a manner of services composition that fits for cloudmanufacturing environment is to be of important theoretical and realistic significancefor the emerging cloud manufacturing. In view of high dynamic and adaptive features ofcloud manufacturing, this paper studies on issues of manufacturing cloud servicescomposition and proposes the corresponding solution, which combines with semanticinformation and Hierarchical Task Network(HTN) Planning algorithm.Initially, paper sets forth the concept and features of cloud manufacturing andintroduces corresponding knowledge about semantics and ontology. After analyzing andcomparing all kinds of traditional services composition manners, we get the conclusionthat traditional services composition manners are not fit for environment of cloudmanufacturing and efficient, automatic composition manners may be the futuredevelopment direction. Automatic composition can not do without semantic informationand ontology. There is no unified and standard ontology model for cloud manufacturing,although semantic and ontology technology have always been developing. This papergives a concept model of cloud manufacturing ontology by integrating incurrentstandards, which can support automatic composition and provide reference for futureresearching on cloud manufacturing ontology standard.Classic HTN which generates only one plan can not give users more options andthe platform needs to plan again if the only one plan executes failure. In view of theseproblems, this paper extends traditional HTN. Our method includes two parts. First,using utility values which calculates by non functional properties(NFP) modifyplanning process. Second, using control policy based on threshold decreases planningspace. Then, to enhance the stability and practicability of extended algorithm, we improves extended HTN by calculating utility values based on context environment andadjusting control policy. Average method and Markovian Decision Process(MDP)model are built to evaluate multi plans generated by extended HTN. Multi plans cangive users more options and increase the robustness of cloud manufacturing platform,plans evaluation can be as reference when uses choose plan.Finally, based on the study above, realize the algorithm and develop a prototypesystem to verify availability of extended algorithm.Extended HTN can meet the multi plans demand of manufacturing cloud servicescomposition. The algorithm is stable and practical, which can provide reference forfuture research.