Research On Resource Allocation And Computation Offloading Based On Mobile Edge Platooning Cloud In Vehicular Networks

As the global automobile industry evolves,issues like traffic congestion,driving safety,and environmental pollution increasingly prominent.To meet the diversified demands of driving comfort,safety,and environmental friendliness in modern transportation,the Intelligent Transportation System(ITS)has emerged.Leveraging technologies like Vehicle-toEverything(V2X)and Intelligent and Connected Vehicles(ICVs),this system establishes an efficient,intelligent,and fully interconnected network to enable information exchange and data transmission among vehicles,infrastructure,pedestrians,and the cloud.Nevertheless,while advancing the deep development of ITS,Single-Vehicle Intelligence(SVI)displays clear constraints in data storage,processing,and analysis.Vehicular Edge Computing(VEC),a critical component of the intelligent transportation domain,improves data processing efficiency by descending computing resources to vehicles or their vicinity.Nonetheless,as this computing paradigm proliferates and finds application,it confronts issues related to constrained resources,elevated mobility,sporadic connectivity,and scalable service.Therefore,it is imperative to deeply optimize resource nodes within vehicular networks,explore new interaction strategies,and integrate cutting-edge technologies like blockchain and artificial intelligence to improve the overall performance and stability of the system.Recently,vehicle platooning has emerged as a novel paradigm in automated driving,offering transformative opportunities for ITS and VEC due to its distinctive driving patterns and abundant resource potential.Consequently,understanding how to efficiently harness and optimize platoon resources to further the sustainable growth of ITS and VEC is an imperative and timely topic for study and research.This paper presents strategic solutions to the outlined challenges,segmented into four pivotal components: the construction of the Mobile Edge Platooning Cloud(MEPC),an analysis of singular static scenarios,an examination of dynamic cross-platoon contexts,and the creation of a market-centric service approach.In the initial segment,the systematic construction of the MEPC is explored.Drawing on the foundational concepts of the MEPC,the subsequent two segments emphasize strategies for resource allocation and computation offloading,aiming to enhance the MEPC’s operational efficiency.The concluding segment,rooted in the service perspective,lays the groundwork for a comprehensive trading marketplace.Throughout this work,the four research facets are intricately linked,forging a cohesive and thorough research trajectory that spans from the inception of the MEPC and technical support of resource strategies,culminating in the establishment of a service trading platform.(1)Mobile Edge Platooning Cloud-A Lightweight Cloud in Vehicular Networks: In the opening segment,this paper delves into the construction of the MEPC,offering a detailed analysis of its synergistic interactions with cloud,air,edge,and end.Recognizing the resources and distinct interaction features of platoons,such as robust V2 V communication,sustained mobile topology services,and the augmentation of edge services in distant areas,a cloud service architecture tailored for platoons in autonomous driving mode is introduced.This addresses performance constraints observed in SVI and the challenges inherent to VEC.Through the incorporation of blockchain and AI technologies,the traditional cloud service architecture undergoes a profound enhancement with the MEPC.This novel approach markedly bolsters the collaborative efficiency of the MEPC across various application contexts,leading to a notable uplift in overall system performance.(2)Consortium Blockchain-Based Computation Offloading Using Mobile Edge Platoon Cloud in the Internet of Vehicles: In the context of MEPC-focused research,the second segment of this paper delves into efficient resource allocation via the service pricing mechanism within a singular static scenario.This study combines the cost-profit model with the Stackelberg game to address the gap between the computational demands of task vehicles and their hardware and software capabilities in complex application environments.This combination aims to derive an optimal resource allocation scheme,enhancing the collaborative efficiency between the task vehicle and the MEPC.Further bolstering data security,the integration of coalition blockchain technology with the Raft consensus mechanism is explored.By synthesizing these technological strategies,the research significantly amplifies the resource efficiency of the MEPC and ensures service security stability.(3)SFO: An Adaptive Task Scheduling based on Incentive Fleet Formation and Metrizable Resource Orchestration for Autonomous Vehicle Platooning: In light of the growing traction of MEPC implementations,the third segment of this paper crafts an adaptive task-scheduling strategy tailored for intricate application contexts.Moving beyond the confines of unidirectional driving formation maneuvers,an innovative formation algorithm is presented.This algorithm,rooted in vehicle trajectory matching and joining willingness,seeks to foster prolonged,stable collaborations within MEPCs.Concurrently,to reduce time-averaged energy consumption,a resource scheduling approach is designed.This approach considers factors such as application divisibility,fluctuating resources,and execution latency,aiming for a near-optimal solution based on the changing backlog of timeout queues.This comprehensive strategy augments both the stability and service efficacy of the MEPC.(4)Multi-Agent Reinforcement Learning based Trading Decision-making in PlatooningAssisted Vehicular Networks: After a thorough discussion of MEPCs and their diverse application scenarios,the fourth segment of this paper focuses on the construction and optimization of the resource trading market,advancing the market-oriented implementation of MEPCs.To ensure high accessibility of services and further maximize the long-term mutual benefits between MEPCs and task vehicles,the entire trading process is modeled as a multi-objective optimization challenge,simulating the dynamics and uncertainties of realworld settings.Given the challenges of multi-objective optimization in the trading process,this paper designs a Global-Local training architecture,incorporating a Hybrid action space and Prioritized sampling into a Multi-agent reinforcement learning algorithm that utilizes a Twin Delayed deep deterministic gradient(GL-HPMATD3).This facilitates rapid decisions related to service selection,resource allocation,and trading pricing.The overall layout of the service trading strategy propels the in-depth market-oriented development and practical application of MEPC technology.This strategy not only spawns a mutually beneficial trading model among market participants but also ensures the efficient execution capability of real-time decisions during the trading process.