Research On Task Offloading And Resource Allocation For Edge Computing Of Internet Of Vehicles

With the rapid development of Internet of vehicles technology,the applications on vehicle devices tend to be computationintensive and timedelay sensitive,which brings significant pressure to onboard devices with limited computing resource and cache resource.Edge computing allows onboard devices to access computing and cache resource on close edge servers by utilizing task offloading technology,which brings a new approach to address the issue of insufficient computing capacity of onboard devices.Nevertheless,the system complexity and resource capacity limitation of edge computing in Internet of vehicles pose challenges to the efficient utilization of resource and the guarantee of user service quality.To effectively utilize the resources of the Internet of vehicles,reduce the task implementation delay and improve system efficiency,this paper studies task offloading and resource allocation in edge computing of Internet of vehicles.The main research contents are as follows:Firstly,in order to address the issue that onboard devices are difficult to meet the computing requirements of delaysensitive applications on the Internet of vehicles,a delayoriented task offloading strategy of edge computing for the Internet of vehicles is proposed.In the edge computing,the task offloading problem for the Internet of vehicles is transformed to a delay optimization problem with resource constraints,task divisibility and task serialparallel timing constraints.Considering the dynamic change of Internet of vehicles and the numerous constraints,the optimization problem is constructed as the Markov decision process.The state,action and reward function are designed according to the task features and resource limitation.Then the deep reinforcement learning is utilized to solve it.The proposed offloading strategy minimizes the delay cost of the edge computing of Internet of vehicles.Secondly,in order to solve the problem of insufficient computing resource of a single edge server in vehicledense areas,a resource allocation optimization strategy based on edge server cooperation is proposed to schedule the idle resource of nearby edge servers.The cooperative resource allocation problem is formulated as the system efficiency maximization problem.By taking each edge server as one agent,the cooperation strategy is obtained based on the multiagent deep deterministic policy gradient algorithm,and the strategy is decomposed into centralized training and distributed execution stages.The central cloud delivers the trained optimal models to edge servers,which then optimize resource allocation according to task features,effectively reducing the implementation time of tasks and improving the overall system efficiency of edge computing of the Internet of vehicles.The simulation platform of the edge computing is constructed for Internet of vehicles.A numerous tests are conducted to verify the effectiveness of the task offloading and cooperative resource allocation strategy on reducing delay cost and improving efficiency.