Research Of Computing Offloading Scheme For MEC-enabled Vehicular Networks

With the dramatic development of automobile manufacturing and the Internet of Things,many new types of vehicular applications have emerged to provide many conveniences for people.These applications require enormous storage and computation resources to process related data.And these applications need higher requirements for the quality of service in network.The existing vehicular network will not be able to meet the requirements of these new vehicular applications in computation resources and latency constrains.To solve this problem,mobile edge computing(MEC)is introduced into the vehicular networks.Nowadays,the MEC-enabled vehicular networks have gained extensive attention from academia and industry community.The mobility of vehicles will have a great impact on the MEC offloading and results reply transmissions.It makes the network scenario and resources allocation becoming more complex.Thus,it brings many new problems to be solved.MEC servers are generally deployed on the road side units.And their service range will be affected by the wireless communication range of road side units.Thus,the mobility of vehicles may result in interruptions in the process of computation task offloading.In addition,the computation tasks can be offloaded to other vehicles with idle computation resources by the vehicle to vehicle communication.How to make a choice from various offloading methods is also a worthy problem to study.And the main work of this paper can be introduced as follows:In order to avoid the interruptions in the process of offloading from vehicles to MEC servers,a dynamic offloading scheme to avoid the interruptions is proposed in this paper.In the system model,many practical influence factors are considered,including the dynamic change of vehicle speed,the coverage of cell,the transmission data rate,computation resources and so on.And the offloading is dynamic changing according to these factors.The scheme can minimize the latency under the promise of no interruption in the offloading process.In order to improve the performance of adapting to speed changes,the computation task is divided into many small task units.Each small task unit can be either computed locally,or offloaded to the MEC server for computing.Then,considering the real-time vehicle speed,bandwidth,communication channel,computation resources and some other factors,a suitable pre-allocation scheme of task units is proposed.Finally,these pre-allocated task units are offloaded by partial offloading.And the optimal ratio of the number of task units offloaded to the MEC server to the total number of pre-allocated task units in a single cell can be calculated.The simulation results show that the proposed dynamic offloading scheme have a good performance in latency and energy consumption.And it can effectively avoid the interruptions in the process of offloading.In this paper,a federated offloading scheme is proposed to solve the problem of how to efficiently complete a computation task with multiple offloading methods.Considering the computation resources of local devices,collaborative vehicles and MEC servers,we investigate the problem and find an optimal of:floading scheme.The scheme achieves the goal of minimizing the total delay of offloading,and can improve the utilization of resources.Many practical constrains in the vehicular scenarios are considered,including relative speed between vehicles,irregular distribution of computing resources,bandwidth,path loss and some other factors.Firstly,the computation task is divided into three parts,which correspond to the above three types of computation resources respectively.Secondly,by analyzing the impact of the offloading sequence on the total latency,two federated offloading modes are proposed.Under different environment conditions,the performance of the two federated offloading modes is different,and the optimal one is chosen as the federated offloading scheme.In addition,a distributed algorithm for offloading between collaborative vehicles is proposed,which can find an optimal offloading routing between collaborative vehicles according to the topology of the vehicular networks and the input computing task size.The simulation results show that the proposed federated offloading can effectively reduce the latency for completing the computation task and improve the utilization of computation resources.