Research On Communication Membrane Calculus And Its Applications Research On Communication Membrane Calculus And Its Applications

Natural computation is a series of computational models and algorithms that simulate structures,functions or hierarchical relationships of nature.According to sources of inspiration about computational models,natural computation can be divided into physical hierarchy,life hierarchy and cultural hierarchy.In the field of computer science,genetic algorithm,ant colony algorithm,neural network and so on belong to the life hierarchy of natural computation.As a branch of natural computation,membrane computing has developed rapidly in recent years.In this paper,based on the theory of membrane computing,two new variants of membrane computing are proposed and their applications are studied.When studying problems of service composition and software product line,it is found that the existing membrane computing theory is not suitable for solving problems about isolated island in service composition and evolution of software product line.With regard to the isolated island problem,there should be communication between an island and a system,that is,communication between a system and its external environment.And in order to deal with evolution of software product line,the process of adding new change mechanisms into existing systems needs to be implemented.Existing membrane computing models can not solve the above problems.Therefore,based on the existing theoretical research on membrane computing,this paper proposes two new membrane calculus,channel membrane calculus and communication membrane calculus.They are suitable for solving isolated island problem in service composition suitable and modeling evolution of software product line,respectively.The main contributions of this paper as follows.(1)Current membrane computing models suppose that,the transfer of the object and the membrane must be in the same surface membrane.This definition indicates that they are not good at describing communication between two different environments.Therefore,in order to deal with the above situation,a new computing method called channel membrane calculus is proposed.This method extends reaction rules of membrane calculus,and defines a way of passing information using transfers among objects and membranes through channels and membranes.In the definition,the creation of a channel is not limited to the same surface membrane,so different surface membranes can pass information(i.e., objects and membranes)through channels.This method is suitable for modeling isolated island problem.(2)In the existing membrane computing models,it is assumed that objects and membranes should change according to reaction rules.This definition indicates that reaction rules of these moddels can not be changed,so they are not good at modeling systems that change in an unfixed way.Therefore,a new variant of membrane computing,called communication membrane calculus,is proposed. Based on definitions of the original membrane computing model,communication membrane calculus allows rules to be transferred,replicated and deleted,which provides reaction rules with the capability of dynamic change.Therefore, communication membrane calculus has more mechanisms of communication and change than the existing membrane computing models.Communication membrane calculus can be used to model evolution of systems,such as evolution of software product lines.(3)Communication membrane calculus and channel membrane calculus proposed in this paper are compared with related membrane computing models.By using concepts of state,successive state and arrival path,several computing models are compared in detail from the aspects of static structure description capability and dynamic evolution ability.The comparison results show that,channel membrane calculus is suitable for static structure description,and communication membrane calculus does well in the modeling of dynamic evolution problems.(4)Traditional methods for evolving feature models of software product lines mostly rely on analysis of domain experts.Once software product line has changed, domain experts need to reconstruct its feature model.This reconstruction process is completely dependent on the domain experts' understanding of software product line and evolutionary requirements,which can not guarantee correctness of the feature model.Therefore,based on evolutionary requirements,this paper proposes an automated generation method for feature model of software product lines,and structure and evolution of feature model are described using communication membrane calculus.The method proposed in this paper can automatically determine the evolution range of a software product line,automatically change its feature model,reduce the workload of domain experts during feature model evolution and avoid errors caused by manual correction of feature model.