Task Offloading For Autonomous Vehicle Based On Combinatorial Auction

With the rapid development of autonomous driving technology,the number of computingintensive and latency-sensitive applications on autonomous vehicles is rapidly growing and attracting extensive attention.The limited computing power on the vehicle is gradually unable to meet the increasing computing demand,and the task offloading technology under edge computing emerges as an effective solution.In order to improve the performance of autonomous vehicles and enhance the user’s driving experience,it is the main problem of current research to formulate a reasonable offloading strategy and efficiently allocate the resources of edge computing nodes during task offloading.Edge computing nodes have limited resources and cannot provide computing services for all vehicles at the same time,and will not provide computing services for other vehicles when it is unprofitable.At the same time,each vehicle is a selfish individual with a conflict of interest in resource allocation.Therefore,vehicle task offloading under edge computing faces complex resource allocation problems.Many studies only optimize resource allocation with the goal of system delay and energy consumption,while ignoring the impact of economic benefits on resource allocation of edge computing nodes.That is,these task offloading schemes lack incentives for vehicles and edge computing nodes.At the same time,most studies consider resource allocation in a few dimensions,such as computing resources,storage resources,etc.And they do not consider the allocation of more types of resources.To sum up,how to efficiently allocate various types of resources of edge computing nodes and coordinate the conflicts of interest between different entities is the main problem to be solved in this paper.Combinatorial auction can efficiently allocate multiple resources,and when these resources are complementary or substitutable,it can achieve good economic benefits.Therefore,the goal of this paper is to propose a vehicle task offloading scheme based on combinatorial auction theory.In view of the above problems,the research content of this paper includes the following aspects:First,based on the vehicle task offloading architecture under ultra-dense networks,this paper proposes a vehicle task unloading scheme in ultra-dense networks based on combinatorial auction.Assuming that all edge servers are provided by the same service provider,vehicles acquire the required resources through a combined auction.In the ultra-dense network architecture,the control layer acts as the auctioneer and the vehicle acts as the bidder.The vehicle initiates a bid for the required resource combination,and the control layer collects global information,including bidding information and remaining information of edge computing node resources,and allocates resources with the goal of maximizing the total revenue of the edge server.In the mentioned auction scheme,the winner determination algorithm and payment determination algorithm based on greedy strategy are designed,and it is proved that the auction scheme satisfies individual rationality,computational efficiency and authenticity.Secondly,considering the heterogeneity of edge servers,this paper proposes a double combinatorial auction task offloading scheme for heterogeneous edge servers.The vehicle task offloading under heterogeneous edge servers is divided into two stages,local auction and double auction.In the local auction,the edge server allocates resources to the vehicles under the server through the combinatorial auction.When there are still tasks that cannot be successfully offloaded after the local auction,the server can initiate a double auction.Edge servers offload local tasks to other edge servers through double auctions,which can improve the resource utilization of the entire system and improve the economic benefits of edge servers.In the auction scheme,the greedy strategy and the Mc Afee double auction algorithm are combined,and the winner determination algorithm and payment determination algorithm for solving the double combinatorial auction are proposed,and it is proved that the auction scheme satisfies individual rationality,computational efficiency and authenticity.Finally,the proposed scheme is simulated and tested in this paper.The experimental results show that both schemes can improve the utilization of system resources and increase the revenue of edge servers.Through the analysis of auction efficiency,it is found that the vehicle task offloading scheme based on ultra-dense network is suitable for delay-insensitive vehicle tasks,while the double auction task offloading scheme based on heterogeneous edge servers is suitable for delay-sensitive vehicle tasks.