Research On Intelligent Service Selection Algorithm For Ubiquitous Computing

Ubiquitous computing makes it possible to integrate the physical world we are living in and the virtual world in the information space together as the whole,where the users can expediently and transparently obtain digital services.It would enable people to focus on their task itself,instead of the computer,so as to make the computer invisible from users' sight.Therefore, it is necessary to provide a kind of mechanism that users can find and select devices or services.The process must be transparently completed to users, which is also one of the main goals for ubiquitous computing.In ubiquitous computing system,because of the increasing of computing complexity and mobility,there must be more than one device to cooperate to finish the target task.So how to choose a suitable service from all the useable service candidates is the most important step for ubiquitous computing.For the problem that in traditional computing model the users only select service according to their function of services,the relation between user and service provider is close coupling and it didn't consider that context information and other correlative factor severely affect service selection activity.There is not an effective guidance provided to users,so the selecting process is blindly and arbitrarily implemented.Its can be the results of low system performance and bad efficiency.In this thesis, beginning from the basics of service discovery and selection,we research the intelligent method to select the optimum service and migrate the failure service.Then,we put forward a novel ANN-based intelligent service selection model(ANNSSM) and algorithm(ANNSSBP) for ubiquitous computing;furthermore,we propose a novel service-oriented seamless migration(SOSM) algorithm for ubiquitous computing.We evaluate performance and effort of proposed project by adopting method of theory analysis and simulation.Finally,we develop a prototype system to testify the validity of proposed algorithms.The main contents and contributions of our work are as following:Firstly,we put forward the architecture of ANN-based(Artificial Neural Networks) intelligent service selection system for ubiquitous computing,and propose a novel ANN-based intelligent service selection algorithm for ubiquitous computing.This algorithm is based on the earlier knowledge of information of the cooperation among devices and the context information of ubiquitous computing environment.At the same time,in our novel model,the Neural Networks Prediction Controller is adopted,and an evaluation about the target service is given out according to the properties of the services.Consider that the evaluation value can give out a guidance standard,so user can choose a most suitable service from many service candidates.The suitable service can provide the most perfect performance to user.Because of adopting the novel ANNSSBP algorithm,the process of service selecting is intelligentized,and it makes that the relation between user and service provider is changed from close coupling to loose coupling. By adopting this kind of self-adapting architecture of service selection system,the system efficiency can be observably improved and the Quality of Service(QoS) is observably improved.Secondly,we put forward the novel method of self-adjusting architecture of Artificial Neural Networks,and improve BP algorithm based on three-term(namely Learning Rate,Momentum Factor and Proportional Factor,shortly LR,MF and PF).In order to optimizing the architecture of Artificial Neural Networks,the self-adjusting architecture method is adopted by uniting or cutting down nodes of network so that a moderate scale of neural network can be obtained.At the same time,we adapt the method of proportinal factor among all kind of correlative parameters to avoid tripping into local minimum in the course of iterative process.The convergence speed and stability of improved BP algorithm were enhanced by adding the proportional factor,which make it possible to satisfy the new characteristic of ubiquitous computing.Thirdly,we put forward a new service-oriented seamless migration algorithm for all kind of failure service in ubiquitous computing environment,and propose the condition and judgement theorem of triggering service migration active,judgement theorem of successful service migration active.We research on some correlative problems and give out some simulations,including rate of correctly resuming for service checkpoint,rate of being gratified for migration delay,rate of remains dependent for service restarting and so on.One of these problems what we detailedly research is the mechanism of optimizing service migration route. The new SOSM algorithm can make service seamlessly migrate among the nodes of ubiquitous computing networks and optimize the global computing resource,which enhance the reliability of service system.In a word,this thesis has proposed the novel ANN-based intelligent service selection model and service-oriented seamless migration model, which are new mechanism about service selection and migration from an intelligent-control's point of view.The new mechanism is peculiarly valuable for research and application.Because it fully consider the impact of context information and varying environment,it can enhance the rate of successful service selection active,decrease the blindly and arbitrarily service selecting process,and improve the system reliability so that system be optimizing.It is really valuable to research and develop intelligent service selection system in order to develope ubiquitous computing research.This work is partially supported by the Innovational Thesis of Doctor Degree of Donghua University and the Ministry of Education Technology Research Key Foundation of China under grant.