Intelligent Resource Scheduling Method Of Vehicular Edge Collaboration Network Under Uncertain Conditions

With the development of Multi-access Edge Computing(MEC)and Internet of Vehicles(Internet of Vehicle,IoV),vehicles can be used as computing nodes which can provide computing resources with server together.This is called Vehicular Edge Collaboration Computing(VECC).However,in VECC,the high-speed movement of vehicles causes uncertainty in the network topology and network resources.At the same time,most of tasks in VECC are computationally intensive and time-sensitive.The computing resources of a single vehicle cannot afford to the computation of such tasks.Therefore,how to reasonably allocate resource in VECC has become a new challenge.The resource scheduling method in VECC can be divided into two parts:computing offloading and service migration.Computing offloading allows complex tasks to be offloaded to other nodes under some constraints,such as delay.Most research lacks the integration of heterogeneous resources in VECC.At the same time,research on dynamic delay constraints is not sufficient.Service migration aims to migrating the whole running services to other nodes which should ensure the continuity and quality of service.Currently,there are many problems,such as migration delay and repeated migration.For overcoming the shortcomings in existing research,this article studies the resource scheduling scheme in VECC from the following two steps:(1)We propose an intelligent joint offloading algorithm based on the dynamic constraints of environment.First,we use the vehicle’s sensors to establish mobile crowd sensing network(MCS)which is aimed at predicting the speed and density in the next time slot based on the vehicle’s current traffic conditions.Then we calculate the vehicle’s communicable time.Second,we integrate the resource conditions of heterogeneous nodes such as vehicles and edge servers.We fully consider the dynamic delay constraints of the nodes,and establish an offloading scheme based on the deep reinforcement learning algorithm to minimize the offloading delay and realize the reasonable allocation of resources.Finally,simulation experiments prove that the proposed algorithm has good convergence.It can accurately predict the traffic network status,and effectively reduce the offloading delay.(2)We propose a self-learning migration algorithm for delay prediction of VECC.In order to solve the migration delay,we first predict the resource demand of each node in VECC.Due to the large amount of data in VECC,the sliding window mechanism is adopted to ensure the accuracy of this model.Then we establish a forward-looking service migration algorithm based on deep reinforcement learning.We fully consider the available resources of each node in the next time slot which can improve the success rate of migration,and avoid additional cost caused by secondary migration.Finally,simulation experiments prove that the proposed method has good convergence,high accuracy,high migration success rate.This algorithm effectively avoids secondary migration,and realizes reasonable resource allocation.