Research On Task Scheduling And Pricing For Edge Computing In Autonomous Driving

With the improvement of the performance of autonomous vehicles and the development of intelligence,various autonomous driving applications are growing rapidly.The increase in on-vehicle applications has brought explosive computation demands to autonomous vehicles with limited onboard resources.Edge computing can effectively alleviate this problem through application offloading.Some on-vehicle applications such as speech recognition,e-mail,interactive games and so on can be offloaded to the edge server for execution.Because edge computing nodes are close to vehicle terminals,they can provide vehicle users with low-latency and highly reliable services.Nonetheless,due to the limited computing resources,edge servers cannot meeting the latency requirements for all offloaded computing tasks,resulting in an inevitable resource contention between vehicles.The latency sensitivity and pricing of different in-vehicle computing tasks are different.In this regard,how to dynamically and reasonably allocate computing resources to improve resource allocation efficiency and social benefits is a big problem for edge servers with limited resources.In response to the above problems,this article has conducted in-depth research,and proposed corresponding computing offloading schemes and resource pricing strategies.The main content of this article is as follows:(1)Aiming at the scenario where multiple edge servers distributed within a certain road range belong to the same computing resource provider,this paper proposes a centralized scheduling strategy for vehicle task offloading.According to the characteristics that the central server collects the vehicle task information and updates the edge server information in real time,we propose an extended stable marriage matching algorithm to match the vehicle task and the edge server.Considering the utility of server providers and vehicle users,we establish vehicle task preference list related to server revenue and edge server preference list related to task service quality.According to the weight of different tasks and the utilization of computing resources,we propose a task screening algorithm to solve the problem of excessive load on the same edge server.Finally,we propose a task offloading evaluation model based on modern portfolio theory to balance the benefits of edge servers and the risk of vehicle task delay.The simulation results show that compared with other heuristic scheduling algorithms,the proposed strategy can increase the overall system revenue by 63%on average,reduce the risk of task overrun delay by 70%on average,and the server load is balanced.(2)Aiming at the scenario where edge servers distributed within a certain road range belong to different server providers,this paper proposes a distributed pricing strategy for vehicle task offloading.Taking into account the problem of resource contention between vehicle users and revenue competition between edge service providers,we use the Stackelberg game model of multi-master and multi-slave form to model the process of vehicle task offloading and resource allocation under edge computing.Considering that the edge servers are isolated from each other,the server independently adjusts the price strategy based on the interactive information with the vehicle user.Through price adjustment,the edge server influences the purchase choice of vehicle users’ computing resources,and finally obtains the best price strategy under the current traffic flow.We propose a task selection algorithm based on greedy thinking,which solves the problem of excessive server load under each pricing,and achieves the optimal server utility under this pricing strategy.Finally,the experimental results show that the distributed strategy proposed in this paper requires an average of 250 iterations in urban scenarios,and an average of 150 iterations in high-speed scenarios.The algorithm has good convergence.(3)In this paper,the computational offload scheme and resource pricing strategy are simulated.The results show that under the centralized scheduling strategy of vehicle task offloading,the overall system benefit is high,the task risk is small,and the server load is balanced.This scheduling strategy can increase the revenue of the server provider while ensuring the quality of user service.The experimental results also show that the distributed offloading strategy can effectively solve the multi-competition problem and has good convergence.This distributed pricing strategy can make full use of server resources while meeting the greater utility of edge servers and vehicle users.In summary,the strategies proposed in this article can guarantee the efficiency of resource allocation and social benefits when edge resources are limited.