Reasearch On Mobile Edge Computing Service Composition Mechanism Based On Deep Reinforcement Learning

As the core technology of business process,service composition has been widely concerned by academic.With the rapid popularization of 5th generation mobile network,new intelligent applications have the characteristics of complex and diverse,long duration,and large amount of data.However,traditional cloud service composition technology can not meet the current real-time business requirements.Mobile edge computing can reduce transmission delay and network congestion by sinking services and resources to the edge of the network.The application of MEC in service composition can reduce the load of cloud and improve efficiency.Although some progress has been made in the research of service composition based on mobile edge computing,related work has not considered the new challenges brought by the mobility of intelligent terminals to service composition.Some studies have not consider the problem that change with the geographical location of the smart terminal,resulting in the failure of result delivery.Part of the work is inefficient when deep reinforcement learning is used to deal with multi-dimensional service resources in a dynamic network environment.Therefore,this thesis studies the service composition problem based on mobile edge computing.The main work is as follows.1.For the service composition problem of mobile edge computing,it is divided into two phases,including the service discovery phase and the real-time service composition phase.The service discovery phase filters failed and redundant services and recording the dependencies between services,a dynamic service network structure is constructed.In the real-time service composition phase based on reverse exploration,the reverse search is carried out according to the dynamic service network,and the next executable service is determined by comprehensively evaluating the quality of the service,and finally optimal composition result is obtained.The experimental results show that the proposed algorithm has better efficiency and higher success rate than existing classicial algorithms.2.For the high-dimensional dynamic service composition problem,the request task are divided firstly.Secondly,it is modeled as a Markov decision process model.Thirdly,the service migration strategy solves the problem that the intelligent terminal fails to deliver the results due to the high-speed movement.the deep reinforcement learning algorithm is used to solve the data explosion problem,and the Q value is estimated by the neural network for service selection.At the same time,the structure of the neural network is optimized to solve the problem of prediction Q value too large.Finally,it is verified by simulation experiment.The results show that the algorithm proposed in this thesis is more adaptable to changes of the network environment.Compared with existing algorithms,it has obvious advantages in terms of efficiency and quality of service composition.In summary,service composition in mobile edge computing environment is studied.Different solutions are proposed for different business needs and application scenarios,which provides a theoretical basis for further exploration and wide application in this field.