Research On Offloading Mechanism Of UAV-Assisted Vehicular Edge Computing

In recent years,with the rapid development of the Internet of Vehicles and intelligent transportation system.Latency-sensitive and computationally-intensive vehicular applications(such as infotainment,autonomous driving,route planning,etc.)are emerging.This has brought great challenges to some intelligent vehicles with limited on-board storage capacity and computing resources.Although using Vehicular Edge Computing(VEC)to offload the tasks generated by the vehicle terminal to the edge side can effectively reduce the computing pressure of the vehicle,but the edge servers in the traditional VEC network are usually deployed on the roadside infrastructure,which leads to its shortcomings such as limited coverage and low flexibility,and it is difficult to adapt to dynamically changing traffic density and vehicle topology in different time periods.Due to the high mobility and easy deployment of unmanned aerial vehicle(UAV),research on UAV-assisted edge computing system has become a new hot spot in recent years.In this system,UAV provides computing service for ground users by carrying server,which effectively makes up for the shortcomings of traditional edge computing.However,in practical applications,UAVs are often unable to provide services for users for a long time due to the limited energy of the onboard batteries.In addition,most of the existing research work often assumes that the ground users are in a stationary state,and the proposed offloading mechanism is difficult to apply to the edge computing scenarios of the Internet of Vehicles with fast moving topology.Based on the above analysis,this paper studies the problem of maximizing system computation efficiency under UAV-assisted vehicular edge computing,which simultaneously measures the number of computation bits and energy consumption of the edge computing system.Considering the limited computing resources of a single UAV,this paper proposes a double-layer UAV architecture to assist vehicle users to complete the computing tasks.This architecture makes full use of the large coverage of the higher-layer UAV,which can communicate with distant base station through line of sight channel.It also utilizes the maneuverability of the lower-layer UAV to provide effective service for hotspots to make up for the lack of single UAV,thereby meeting the vehicle’s requirements for task completion delays.The main research contributions of this paper are as follows:(1)For the one-way road scenario,a vehicle users computing offloading mechanism is proposed,which jointly optimizes computing task offloading scheduling,system resource allocation,and double-layer UAV position deployment,thereby maximizing the computation efficiency of the double-layer UAV system.Then this paper constructs the corresponding optimization problem based on this mechanism.In this optimization problem,the task offloading mode is binary offloading mode.Due to the existence of non-convex constraints and objective function,there is a coupling relationship between different optimization variables.The objective optimization problem becomes complex and hard to solve.Therefore,this paper first decomposes the original problem,and then transforms the sub-problems by using methods such as Successive convex approximation(SCA),and finally proposes an iterative solution algorithm.The simulation results show that the proposed algorithm can converge quickly,and compared with other benchmark algorithms,this algorithm can achieve higher computation efficiency.(2)For intersection scenarios,a multi-element joint optimization scheme is proposed in order to maximize the computation efficiency of a double-layer UAV system.In this scheme,this paper comprehensively considers the influence of factors such as vehicle task allocation ratio and offloading scheduling,system resource allocation,lower-layer UAV flight trajectory on the system computation efficiency,and then constructs the corresponding optimization problem.In this optimization problem,the offloading mode of computing tasks is a partial offloading mode.This problem is a non-convex optimization problem,which is difficult to solve directly.Considering that the relevant optimization variables have a continuous range of values,this paper transforms the non-convex problem into a Markov decision process(MDP),and then proposes a joint offloading decision algorithm based on Deep Deterministic Policy Gradient(DDPG)to solve the optimization problem.Finally,the simulation results show that the algorithm has good convergence performance,and the system computation efficiency based on the algorithm is better than that of Deep Q-Network(DQN)algorithm and other benchmark schemes.