Research On Computation Offloading In The Internet Of Vehicles

With the advances in technologies such as wireless communication,the Internet of Things,and artificial intelligence,the intelligentization,networking,and electrification of vehicles have also gained great momentum.The intelligent and connected vehicle(ICV)integrates sensing,decision-making,and control functions.In the future,it will realize the transformation of massive information from physical space to information space.The whole process depends on data computation with different fine granularity.Edge computing,as one of the keys to supporting the deployment and landing of the Internet of Vehicles(IoV),can be conducive to processing the massive data generated in the system by providing computing resources at the edge of the IoV.Therefore,vehicles can obtain information far beyond their own perception ability with high efficiency,and realize diversified applications such as object tracking,highly automated driving map(HADMap),multi-level autonomous driving,infotainment systems,etc.The computation offloading is a key research point in vehicular edge computing(VEC).Its solution involves the integration and planning in the field of wireless communication and computing,and the feasibility of the solution is highly related to vehicle driving experience and road safety.However,the limitations of edge node coverage and resources,as well as the high mobility of vehicles,restrict the data offloading throughput and task execution delay of the IoV System.Therefore,this thesis aims to study the solutions of achieving fully optimizing data scheduling and resource allocation based on edge computing and the characteristics of the IoV,and design high-performance computation offloading schemes.The VEC computation offloading schemes mainly face the following problems:first,the problem of redundant data scheduling.The existing schemes mainly focus on improving the efficiency of resource allocation but fail to identify the data characteristics,ignoring the common needs of IoV users for task computing,thus resulting in limited resources serving repeated offloading and computing.The second is the continuous shortage of resources.When vehicles are running or driving for a long time,continuous resources occupation is needed.However,the computing and communication capacities of the edge network are insufficient to support the ever-increasing vehicles.The available schemes put forward three computing ways,including vehicular collaborative computing,unmanned aerial vehicle(UAV)-assisted computing,and satellite computing,but these solutions are challenging to meet the needs of the IoV in terms of security,stability,and reliability.The third is the problem of ineffective computation migration caused by handoff.The handoff of vehicles in the edge network will cause computing interruption.However,the existing schemes focus on task/result migration in the edge network,task re-offloading,and task putting off.These services are relatively lagging and the performance cannot be guaranteed.In response to the above problems,this thesis studies the computation offloading scheme under the IoV,focusing on the data characteristics,the development of resources,and the solutions for the network handoff,to improve the computing offloading performance of the IoV.The specific innovations are summarized as follows:1.In terms of redundant data scheduling,a data redundancy elimination scheduling scheme based on the data characteristics of the IoV is proposed.Compared with the traditional scheme that only considers resource allocation,this scheme pays further attention to the data characteristics of the IoV.According to the current situation that vehicles in the same road section have both common and individual needs in task computing,this scheme is proposed to divide the common data and individual data in the task into two stages for offloading.In the first stage,the multi-vehicle cooperation is applied to the one-time common data offloading.After ensuring the successful common data offloading,in the second stage,each vehicle offloads individual data using available network resources.During the two-stage offloading process,the Non-Orthogonal Multiple Access(NOMA)technique is applied to improve the offloading efficiency.On this basis,the scheme proposes a resource allocation method based on Successive Convex Approximation(SCA).By optimizing time allocation and the user power control within each stage,the individual offloading capability of the worst user is maximized.The simulation results demonstrate that compared with the traditional scheme,the proposed scheme greatly improves the resource utilization efficiency of the communication network,and further improves the data offloading capability of the IoV.2.Considering the continuous shortage of computing resources,this study introduces the high altitude platform station computing(HAPS)into the IoV and proposes a computation offloading framework that integrates vehicular onboard devices,edge nodes,and HAPS.In this framework,the HAPS can also serve as a database of environmental files for the IoV system.Based on the new framework,this study proposes a low-delay data offloading and processing scheme,where the task can be processed at the vehicle,edge node,and HAPS.In addition,this scheme further proposes to deploy the caching function in the edge so as to reduce the transmission delay of environment files from the HAPS to the ground.Aiming to minimize the task execution delay,a decoupling approach based on multi-agent reinforcement learning(MARL)and the Lagrange duality method is proposed to optimize the offloading decision,caching decision,bandwidth allocation,and computing resource allocation.The simulation results indicate that HAPS computing can effectively alleviate the resource shortage problem faced by VEC,and confirm the superiority of the proposed scheme in terms of delay.3.To solve the problem of computing lagging caused by handoff,this study proposes a handoff-aware adaptive computation offloading scheme.This scheme is not limited to computation migration strategies among edge nodes but incorporates parallel computing into the computing framework that is proposed in the last research.According to the diverse needs of users,network dynamics,and the time-sensitivity of handoff,this scheme flexibly divides the tasks of each user into three parts and assigns them to vehicles,edge nodes,and HAPS to perform synchronous computing.In addition,this scheme further constrains the task computing at the edge node to be completed before handoff for avoiding the risk of computing interruption.On this basis,this scheme proposes a data splitting and multi-resource optimization method based on SCA.By jointly planning the task splitting proportion,transmit power,bandwidth allocation,and computing resource allocation,the delay is minimized.The simulation results show that this scheme not only avoids the negative effects caused by handoff in a flexible manner,it also takes into account the delay performance and maintains the delay stability.