Research On MEC-based Offloading Mechanism In Internet Of Vehicles

In emerging vehicular-connected networks,the terminals have more stringent requirements on network bandwidth and transmission delay.The proposal of mobile edge computing(MEC)technology in the research of new communication networks can better solve this challenge.It mainly deploys servers with computing and storage capabilities around the end user,which shortens the distance of data transmission between mobile terminal devices and MEC servers,and reduces the energy consumption of task offloading,and the transmission delay.But in the process of practical application,vehicles users often face the problem of how to select the MEC server,also meet the problem of shortage of channel resources caused by multiple users requesting access.To address these challenges,this thesis mainly makes the following research work:1.Software defined network-vehicle(SDN-V)was introduced into the network system of MEC-based vehicle-connected,and a new task offloading framework was constructed.The SDN-V controller was used to achieve unified scheduling management of global information variables and computing resource data,which ensures the rationality of the task offloading model by obtaining own battery capacity of the battery terminal,combined with multiple constraints such as delay and energy consumption.For the multi-objective optimization model,the Lagrange multiplier method is used to solve the convex optimization model.The diversity of tasks is considered during task offloading.And the concept of "importance" of the task is defined according to the influence factor of the objective optimization function,the task requested by the mobile terminal user are classified,and a task offloading priority algorithm is designed based on the calculation model of the importance to implement an efficient and more logical task offloading strategy.2.Aiming at the problem of the limited channel resources for multi-user task offloading,the thesis introduces non-orthogonal multiple access(NOMA)technology.Compared with traditional orthogonal multiple access technology(OMA),it can serve more users under the condition of same channel resources.A hybrid offloading strategy based on NOMA-MEC is proposed by considering various factors during tasks offloading.The algorithm based on Deep Q-learning Network(DQN)was designed to help vehicle users make channel selection,and the optimal power allocation strategy is obtained by iterative learning of multiple neural networks.According to the analysis of simulation data,the NOMA-MEC-based hybrid offloading strategy proposed in this thesis can effectively optimize the multi-user offloading delay and energy consumption and ensure the user’s benefit to the greatest extent.