Research On Joint Optimization Of Computation Offloading In MEC Network

Mobile edge computing is the key technology to solve and realize 5G vision application.The main feature of mobile edge computing is to push the traditional cloud computing resources,network control and storage to the network edge,so as to better support delay sensitive applications.As the key technology of mobile edge computing,the user decides to offload computing tasks to the server,not only expands the computing capacity of terminal devices,but also reduces the task completion delay and energy consumption of devices.An efficient computing offload solution can improve the quality experience of users,and balance network load.The research results of Mobile Edge Computing in recent years are summarized in this thesis,and make further research on effective computing offloading technology.Edge server computing resources are relatively less than remote cloud resources.If too many users are offloaded,the network load may be unbalanced,and the overhead of users will also be increased.Therefore,under the constraint of the computing resources of the edge server,the thesis designs a task offloading decision-making model based on the multi-objective needs of users,and proposes a power optimization calculation offloading algorithm based on game theory.The classic game theory method is used to solve the decision-making problem of multi-user computing offloading in MEC.There always a set of pure policy Nash equilibrium in the calculation offloading strategy are proved.The transmission delay and energy consumption in the process of task offloading are reduced by the binary search method.The algorithm realizes the Nash equilibrium of multi-user computing offloading game.The simulation results show that the performance of the game-theoretic with power optimization offloading algorithm is better than the traditional game offloading algorithm and the adaptive sequential game offloading algorithm.The game-theoretic with power optimization offloading algorithm proposed in thesis significantly reduces task delay and energy consumption of device,and has superior performance of computing offloading.When a user moves from one server coverage area to another server area,the user needs to connect between the two servers.In this state,in addition to considering the mobility issues of users,the availability of server computing resources must also be considered problem.Therefore,the thesis is based on a single edge server,and for mobility issues of users,the thesis studies the collaborative computing offloading of multiple MEC networks,proposes a joint optimization algorithm of server computing resources based on Q-learning.The thesis defines a core MEC server as the administrator of other servers,and the administrator will allocate certain computing resources to other servers,and supervise their resource usage,and adopts Q-learning The reinforcement learning algorithm updates action and state repeatedly of users.Based on the Bellman equation,the Q function is used to find the best offloading strategy.The Lagrangian method is used to optimize computing resources of server for the offloading computing tasks,and further reducing offloading time of user.The simulation results show that the performance of the joint optimization algorithm of server computing resources based on Q-Learning is better than the random task assignment algorithm,heuristic algorithm and genetic algorithm.The joint optimization algorithm of server computing resources based on Q-Learning proposed in thesis can effectively reduce the task delay and improve the user experience quality.