Research On Event-driven Service Discovery In Semantic Web Of Things

With the popularization and development of Internet of Things(IoT),the intellectualization of Internet of Things has become a hot topic in the industrial community.Applying semantic web technologies to Internet of Things,Semantic Web of Things lays the foundation for the semantic representation and processing of the IoT information,and is considered as an important trend towards the intelligent Internet of Things.In addition,in order to further enhance the interoperability of heterogeneous devices in the intelligent Internet of Things and the capability of agile response,the event-driven service-oriented architecture is applied to the intelligent Internet of Things.Under this framework,device functions are encapsulated in the form of semantic IoT services,providing support for data sharing and interoperability between devices.At the same time,as the carrier of the message,events can reflect the state change of physical environment,and promote the interaction between the intelligent IoT applications and physical environment.When the interesting events in the physical environment are captured,the intelligent IoT applications need to call corresponding services for response processing after the in-depth analysis of the events.Driven by this demand,the accurate identification of events and efficient service discovery become the key and difficult problems.On the one hand,because the environment of Internet of Things is complex and changeable,as well as many unstable factors of networks and devices,there is a certain degree of uncertainty and dynamics for the sensing data required for event recognition.Moreover,in some complex application scenarios of Internet of Things,the use of sensing data is often not enough to comprehensively depict events.On the other hand,eventdriven service discovery needs not only to fully consider the functional requirements of services but also to satisfy the performance requirements as much as possible.From the functional point of view,the description information of events and services should be used to get appropriate services for event response through service matching.From a performance point of view,the quality of service(QoS)should be considered to choose a high-quality service for event response.Therefore,both in theoretical and practical applications,the recognition of events and event-driven service discovery are of great significance and very challenging.This paper focuses on event-driven service discovery in Semantic Internet of Things,and try to solve three key issues which are event recognition,event-driven service matching,and QoS-oriented service selection,in order to promote accurate recognition and agile response of events and provide a strong support for the monitoring of the environment in the intelligent Internet of Things.The main contents and innovations of this paper are summarized as follows:(1)In order to improve the accuracy of event recognition by coping with the limitations of sensing data in event characterization,as well as the uncertainty and dynamics of the data itself,this paper introduces context information into event recognition and proposes an event recognition approach based on Markov logic networks.In this approach,a multi-level information fusion model is proposed.This model adopts the first-order logic formula to construct the relationship among sensor data,context information and events,so that information can be comprehensively utilized.Based on this,the approach deals with the uncertainty information by adding weights to the first-order logic rules.It performs event recognition by using rules attached with weights that are obtained by the statistical learning method and an adaptive update algorithm for weights proposed in this paper.As a result,the approach can weaken the negative impact of data uncertainty and dynamics on event recognition,and improve the accuracy of event recognition in complex IoT environments.The experimental results show that the method of event recognition can effectively utilize the sensing data and context information,deal with the uncertainty and dynamics of data,improve the accuracy of event recognition,and facilitate the environment monitoring and subsequent event response for intelligent IoT applications.(2)In order to improve the efficiency and accuracy of matching events and services,this paper proposes a hybrid training algorithm for word vectors,and then proposes an event-driven semantic IoT service automatic matching method based on word vectors.In this method,two event-related services are defined: event recognition service and event processing service,and then a automatic matching model of semantic IoT service is proposed based on word vectors.The model calculates the matching degree of two services based on word vectors.When the matching degree is higher than a given threshold,it means that there is a matching relationship between the services.In addition,the hybrid algorithm for training word vectors uses the corpus and knowledge base as training materials.At first,it uses Continuous Bag of Words(CBOW)and the proposed Semantic Generation Model(SGM)to process the high-frequency words and the low-frequency words respectively to learning their word vectors.And then,the improved Cosine Similarity Retrofitting(CSR)model is put forward to jointly optimize word vectors for improving the accuracy of service matching by optimizing the quality of the word vector.The experimental results show that the service matching model can effectively deal with eventdriven service matching automatically,and the proposed hybrid training algorithm can obtain high-quality word vectors to improve the accuracy of service matching,which provides a strong support for event agile response in smart IoT applications.(3)Because the function of matched single service is limited,several sub-services can be combined to obtain more powerful composite services.To improve the performance of composite service selection and satisfaction with the quality of service(QoS),this paper takes into account the QoS preferences and constraints,and proposes a method of QoS-oriented composite service selection.The method includes two stages: local optimization and global optimization.In the local optimization stage,the attribute ratio of QoS(equivalent to normalization of attribute Q)and the balance coefficient(used to reflect the influence of candidate service sets on QoS constraints)are proposed.According to attribute ratio and balance coefficient,this approach establishes a local QoS evaluation model that is used to perform QoS evaluation on candidate services.As an evaluation index,this model can improve the performance of composite service selection by reducing the number of candidate services.In global optimization stage,the approach for service selection adopts the fuzzy level analysis to evaluate the QoS preference,and combines QoS constraints to transform the composite service selection problem into an integer programming problem,so as to improve the satisfaction with QoS.The experimental results show that the proposed method fully considers QoS preferences and constraints,improves service satisfaction,and has a high-performance service selection,which helps to improve the response of intelligent IoT applications to events.