Research On Multi-cell Cooperative Cache In Mobile Edge Computing

With the rapid development of technology,mobile device upgrades and people's functional requirements for mobile devices are increasing.Low latency has become an important requirement for improving user's Qo E.Mobile edge computing cache and computing services are pushed to the edge network closer to user to reduce lantency caused by service user requests and offload the backhaul link load.At present,edge caching is a research hotspot in mobile edge computing.However,there is very little consideration of the cooperation between the cells.This paper mainly studies the benefits brought by the cooperation between multiple cells and content is as follows:In this paper,we proposes a multi-cell cellular network model and formulate the multi-cell cache optimization problem as a Stackelberg game problem.The cell control center and the base station group are regarded as the dominant and follower of the game model,respectively,and their respective revenue functions are constructed.Because of their non-continuous function of the income function,the traditional game theory solution can not be used to obtain the Nash equilibrium solution.This paper proposes an iterative alternating algorithm to solve the problem.The control center and the base station group respectively use the improved hybrid frog hopping algorithm(SMSA)and The greedy exchange algorithm(GSA)solves the problem,and the two alternately iterate and finally obtain the approximate solution of the optimal solution of the model.Through numerical simulation experiments,we verify that the proposed algorithm outperforms the greedy algorithm proposed by other researches.In real life,user request data often changes dynamically.In order to adapt to dynamic scenarios to solve cache optimization problems,we consider using reinforcement learning theory for modeling.However,traditional reinforcement learning has defects in some scenarios.In order to make up for reinforcement learning We use the combination of deep learning's strong representation ability and reinforcement learning,that is,deep reinforcement learning method to solve the cache optimization problem.First,we construct the key elements of reinforcement learning and define the update formula of the Q-learning algorithm.In order to avoid the problem of excessive state space of the Q-learning algorithm,we propose a DQN algorithmbased on deep reinforcement learning.Finally,numerical simulation experiments show that the algorithm can effectively reduce the joint cost of delay and return link load.