Research On Edge Computation Offloading Strategy Based On Deep Reinforcement Learning

Edge computing offloading,referred to as computing offloading,is a distributed computing model.By offloading computing tasks from user devices to edge servers,it can improve computing efficiency and response speed while reducing the load on user devices.In the implementation process of edge computing offloading,it mainly includes two stages:computing resource scheduling and communication resource allocation.At present,some studies have used traditional mathematical methods and machine learning algorithms to try to solve the problems in the two stages,but they have not achieved good results.This paper designs an offloading strategy based on deep reinforcement learning,solves the unsolved problems in the two stages of computing offloading,and verifies the effectiveness of the algorithm strategy through experimental simulation.The main work and innovations of this paper are as follows:(1)Aiming at the problem that cloud servers cannot effectively schedule computing resources according to edge server computing resource preferences during computing resource scheduling,this paper proposes a cloud-side computing resource scheduling strategy,which effectively improves cloud-to-edge computing resources.Resource hit rate when scheduling.This paper first establishes the process of cloud and edge collaboration as a personalized recommendation model based on Markov decision process.Secondly,the Actor-Critic framework of deep reinforcement learning is used to construct a computing resource scheduling policy network,and the network parameters are updated using a deep deterministic policy gradient algorithm.Finally,in order to achieve efficient training of the model,this paper designs an online environment simulator for training the policy model in an offline environment.By simulating and comparing the proposed algorithm strategy with some representative schemes,it is proved that the proposed scheme can significantly improve the hit rate of computing resources.(2)Aiming at the problem that the edge base station cannot allocate reasonable communication resources for the user equipment in the process of communication resource allocation,this paper proposes an edge-end communication resource allocation strategy,which effectively solves the problem of time delay,the problem of high energy consumption.This paper first establishes the process of side-end collaboration as a multi-user stochastic game model.Secondly,the Nash-DQN algorithm is used to assign a communication resource allocation strategy for the user equipment.Finally,in order to optimize the solution efficiency of Nash-DQN,this paper designs a local linear quadratic programming method to efficiently solve the Nash equilibrium solution under the multi-user game.By comparing the Nash-DQN algorithm strategy with some currently commonly used schemes,it is proved that the proposed scheme has significant advantages in reducing the delay and energy consumption of user equipment.Under the cooperation of the cloud-edge computing resource scheduling strategy and the edge-end communication resource allocation strategy,the execution efficiency of computation offloading of user equipment has been significantly improved,and the overall system overhead has been further optimized.Through simulation experiments in this paper,it is confirmed that in different system environments,based on the cooperation of the two strategies,the overall delay and energy consumption of the system can be further optimized.