Abstract
With the increasing demand for multiple applications on internet of vehicles.It requires vehicles to carry out multiple computing tasks in real time.However, due to the insufficient computing capability of vehicles themselves,offloading tasks to vehicular edge computing (VEC) servers and allocatingcomputing resources to tasks becomes a challenge. In this paper, a multi taskdigital twin (DT) VEC network is established. By using DT to develop offloadingstrategies and resource allocation strategies for multiple tasks of eachvehicle in a single slot, an optimization problem is constructed. To solve it,we propose a multi-agent reinforcement learning method on the task offloadingand resource allocation. Numerous experiments demonstrate that our method iseffective compared to other benchmark algorithms.