Research On Computation Offloading In Millimeter Wave Train-ground Network

High-speed railway(HSR)is an important infrastructure for national economic development.With the development of wireless communication technology,HSR puts forward higher requirements for train-ground wireless communication.The millimeterwave frequency band has abundant spectrum resources,so broadband millimeter-wave communication is considered to be a promising technology,which can improve the performance of broadband wireless communication services in train-ground communication systems and meet the requirements of high data rates.In addition,FullDuplex(FD)communication has been proposed as a method to improve spectral efficiency.In the HSR train-ground communication scenario,due to the limited battery life and computing power of users,the demand for computing-intensive and delaysensitive intelligent applications is increasing.Therefore,in the HSR scenario,an effective computational offloading scheme in the mm Wave band is necessary.This paper considers an mm Wave-based train-ground communication system in the high-speed railway scenario,where the computation tasks of users can be partially offloaded to the rail-side base station(BS)or the mobile relays(MRs)deployed on the roof of the train.The MRs operate in the full-duplex mode to achieve high spectrum utilization.We formulate the problem of minimizing the average task execution latency of all users,under local device and MRs energy consumption constraints,and propose a joint resource allocation and computation offloading scheme(JRACO)to solve the problem.It consists of a resource allocation and computation offloading(RACO)algorithm and an MR Energy constraint algorithm.RACO utilizes the matching game theory to alternate iterate between two subproblems,i.e.,data segmentation and user association and sub-channel allocation,without considering the constraint of MR energy consumption.With the RACO results,the MR energy constraint algorithm ensures that the MR energy consumption constraint is satisfied.Extensive simulations verify that JRACO can effectively reduce the average latency and increase the number of served users compared to the three baseline scenarios.On the other hand,while the train is running along the track,the MR collects the tasks generated by the user,and forwards these tasks to the BS or the local server for further processing.MR needs to decide where to offload tasks and the proportion of offloaded tasks.But due to changes in the environment and practical requirements,the time and number of arriving tasks is uncertain.We formulate the problem of minimizing the average task execution latency of all users when the unloading user set changes dynamically.In order to solve this problem,we propose a dynamic resource allocation and computation offloading algorithm to minimize the system delay by predicting the delay of users in different association states based on information such as train position and channel.Compared with other benchmark schemes,the simulation results show that the proposed algorithm can maintain the average delay of the system at a low and relatively stable level under different parameters such as self-interference cancellation levels and channel resources.