Research On Edge Computing Task Offloading And Resource Allocation Strategy Based On Deep Reinforcement Learning

With the widespread use of intelligent applications(such as augmented reality,autonomous driving,target tracking,etc.),more and more computationally intensive tasks place higher requirements on the computing power and task processing delay of mobile devices.Mobile edge computing overcomes the shortcomings of insufficient computing power of mobile devices by offloading tasks to the edge cloud closer to the terminal,and at the same time helps to save the energy consumption of the device.Because edge cloud resources are limited,how to choose the location of offloading computing tasks and how much computing resources to allocate for each task is an urgent problem to be solved.In order to solve these problems,a single-cell single-server and multi-cell multi-server oriented edge computing task offloading and resource allocation models for multiple users are constructed respectively.Aiming at the single-cell and single-server scenario,a joint task offloading and resource allocation algorithm based on deep reinforcement learning is proposed.Using adaptive time delay and energy consumption weighting factors based on user experience,a system model with the goal of minimizing task delay and energy consumption of all users is constructed to meet the needs of different users.Transform the task offloading and resource allocation optimization problem into the problem of minimizing system consumption,and define an equivalent form of reinforcement learning,that is,define the state space based on all possible solutions,define the action space based on the computing resource allocation strategy,and propose the concept of termination action to further reduce the action space.On this basis,two effective Q-Learning solving algorithms and Deep Q Network solving algorithms are proposed.The simulation results show that compared with other baseline algorithms,the proposed algorithm reduces the system consumption to the greatest extent.Aiming at the multi-cell and multi-server scenario,with the goal of maximizing system utility,offloading decision-making,communication sub-channel allocation and computing resource allocation are integrated into a mixed integer nonlinear programming problem.Due to the combinatorial nature of the problem,it is difficult to directly solve the optimal solution of a large-scale problem.Therefore,the original problem is decomposed into computing resource allocation,sub-channel allocation and task offloading optimization problems with fixed task offloading decisions.First,the KKT(Karush-Kuhn-Tucker)condition is used to obtain the optimal solution of the computational resource allocation problem,and the subchannel load balancing strategy is formulated in consideration of the channel interference problem.Then,in the task offloading problem,the optimal solution of resource allocation and the sub-channel load balancing strategy are jointly calculated,and a multi-server cooperative task offloading algorithm based on Dueling DDQN is proposed.Finally,through experimental verification,the proposed algorithm can effectively perform computational offloading and resource allocation,thereby improving system utility and reducing the overhead of mobile devices.