Research On Computing Offloading And Resource Allocation Techniques For Edge Networks

In recent years,with the development of 5G/6G communication networks and the emergence of new applications,mobile edge computing(MEC)has become an important technology.It pushes computing,storage,and network resources to the network edge to meet user demands for faster response times and higher energy efficiency.Computation offloading and resource allocation are two key technologies in MEC.Computation offloading refers to transferring computing tasks from mobile devices to MEC servers to reduce the energy consumption and latency of mobile devices.Resource allocation aims to improve system performance and energy efficiency by allocating computing,storage,and network resources reasonably.Therefore,studying the computation offloading and resource allocation problems in MEC is of great significance for improving the service quality and user experience of the mobile Internet.This thesis aims to study the computation offloading and resource allocation problems in mobile edge computing,proposes new solutions and optimization methods to improve the computing performance and energy efficiency of mobile devices.The main research contents are as follows:(1)For optimizing the computation offloading and resource allocation for multiple mobile users in a multi-MEC server scenario,a novel algorithm based on deep reinforcement learning is proposed.Each user in the system is treated as an intelligent agent,and a centralized training and distributed execution approach is adopted.The Critic network of each agent collects the state and action information of all agents for training,while the Actor network of each agent makes decisions for the agent.Through the training of Critic and Actor networks,the algorithm searches for the optimal offloading and resource allocation strategy for each user,and the Noisy Net method is used to optimize the exploration ability of the network model.Experimental results show that the proposed algorithm has better training stability,convergence speed,and optimization performance than baseline algorithms such as DDPG.Moreover,it can reduce system overhead in scenarios with different numbers of users,MEC computing resources,and task computing loads.(2)For optimizing the long-term average system overhead in a multi-MEC server scenario with multiple mobile users and cache-assisted computing,a joint optimization algorithm of computation offloading,local computing resources,edge computing resources,and cache update is proposed.The algorithm first formulates offloading decisions for all users based on their task data size,computational complexity,location,mobility,MEC server available computing resources,and cache content.Then,Genetic Algorithm and KKT methods are used to optimize the local and edge computing resource allocation,respectively.The cache space is updated based on the task request probability,and the approximate optimal decision and long-term average system overhead are learned through the D3 QN network.Simulation results show that the proposed algorithm has better stability,faster convergence speed,and more effective long-term average system overhead reduction under different numbers of users and MEC server computing resource levels,compared with baseline algorithms such as D3 QN and DQN.The use of the cache mechanism based on task request probability can further reduce system overhead and improve system performance.