Research On Key Technologies Of Computation Offloading For Edge Computing In Internet Of Vehicle

With the flourishing development of Internet of Vehicles,vehicular applications,such as real-time navigation and autonomous driving,have emerged,bringing immense computational pressure to vehicular terminals.Edge computing,as an effective solution,alleviates the computational burden of vehicles and satisfies the low-latency requirements of vehicular applications by offloading computation tasks to edge nodes.However,due to the limited resources of edge nodes,it is difficult to balance the trade-off between latency and energy consumption when performing computation offloading in complex vehicular networks.In addition,the rationality of resource allocation directly affects the efficiency of offloading.Improper resource allocation can lead to imbalanced utilization of edge node resources,thereby reducing the service experience.Therefore,key technical research on computation offloading for vehicular edge computing faces the following challenges:(1)for resource allocation,how to achieve a joint optimization of task completion latency and cost,while ensuring a balanced utilization of edge node resources;(2)for computation offloading,how to achieve a balance between latency and energy consumption in computation offloading.To address these challenges,this thesis conducts research on key technical aspects of computation offloading for vehicular edge computing,including:(1)To address the issue of resource imbalance in the vehicular edge computing,a resource allocation method that supports computation offloading is proposed.Specifically,first,models are established for task completion delay,cost,and edge node resource utilization.Then,a resource allocation optimization problem that supports computation offloading is constructed and transformed into a Markov decision process(MDP).Based on the Double Deep Q Network(DDQN)algorithm,a DDQN-based Resource Allocation(DRA)method is implemented for resource allocation in the vehicular network,thereby achieving joint optimization of task completion latency and cost while ensuring the balance of edge node resource utilization.(2)To address the issue of increased latency and energy consumption caused by the complex network state and massive data in the edge computing environment of the vehicular network,a trade-off method between latency and energy consumption for computing offloading is proposed.Firstly,based on the allocated resources of the vehicular edge computing environment,a comprehensive analysis of the computing load status of local and edge devices,as well as the communication status of the system,is conducted to establish models for latency and energy consumption.Secondly,the decision problem for computing offloading in the edge computing environment of the vehicular network is formulated as an MDP.Finally,a TD3-based Computing Offloading(TCO)method is proposed using the Twin Delayed Deep Deterministic Policy Gradient(TD3)algorithm,which achieves a trade-off between latency and energy consumption for computing offloading.