Research On Efficient Offloading Mechanism Of Computing Tasks Based On Intelligent Vehicle Collaboration

The heavy backhaul link load and long response delay make it difficult for the traditional cloud computing architecture to meet the requirements of high bandwidth and low delay for Internet of vehicles applications.However,compared with cloud computing,mobile edge computing（MEC）sinks the computing processing to the edge of the network,which can greatly reduce the computing delay and make it the mainstream solution for the deployment of Internet of vehicles applications.On this basis,in recent years,many research work has begun to focus on how to use vehicles with idle computing resources in the road as service vehicles to assist in the execution of task calculation,forming a powerful supplement to the edge server on the network side.However,the rapid change of network topology caused by high-speed vehicle movement may lead to serious offloading reliability problems and become the main factor restricting the feasibility of vehicle assisted offloading mechanism.Therefore,this thesis focuses on the problem of reliable computing task offloading under the condition of vehicle cooperation.In a large number of existing research work,the guarantee of offloading reliability mainly focuses on how to realize the vehicle collaborative offloading with time delay constraint on the basis of considering the continuity of the communication link between vehicles.On this basis,this thesis effectively expands the service range of vehicle collaborative offloading through RSU relay offloading task or calculation results.This thesis proposes a cooperative vehicle offloading mechanism based on multi-mode joint,which can guarantee the reliability of offloading and and improve the utilization of idle vehicle resources.The mechanism takes RSU as the scheduling center and relay communication node,and optimizes all vehicles covered by the scope as a whole.Four modes are established by combining task offloading mode and calculation results feedback mode.Based on this mechanism,an optimization problem is established to minimize the average completion delay of all task vehicles.The optimization problem considers mode selection,reliability constraints,task offloading ratio and resource allocation.Since the original optimization problem model is a nonlinear integer programming,this thesis decouples it into two sub-problems.Aiming at the problem of task offloading ratio and resource allocation,the convex programming theory is used to transform it into DC（Difference of Convex）problem is solved by the the concave-convex procedure.The genetic algorithm is used to solve the sub-problem of offloading mode decision.Finally,the performance of the proposed mechanism is verified by numerical simulation.On the other hand,this thesis studies cooperative offloading in a more complex vehicle movement scenario,where the duration of communication links between vehicles cannot be accurately obtained.Therefore,this thesis proposes a vehicle cooperative offloading mechanism based on task replication.By probabilistic modeling of the dynamic duration of the inter-vehicle communication link,this thesis applies the idea of task replication to the vehicle collaborative offloading with a fixed subtask segmentation model,and enhances the offloading reliability in highly dynamic scenarios by calculating redundancy.Based on this mechanism,an optimization problem is established to minimize the task completion delay,which takes into account the subtask offloading decision and reliability constraints.Because the original optimization problem model is a nonlinear integer programming,this thesis uses genetic algorithm to solve it.Finally,the effectiveness of the proposed vehicle collaborative offloading mechanism based on task replication is verified by numerical simulation.