Research And System Implementation Of Optimal Control Resource Scheduling Based On Load Prediction In Vehicular Edge Computing

With the rapid development of smart cars and various vehicle applications,concepts such as Intelligent Traffic System and Internet of Vehicles are proposed to solve traffic and driving problems.Technologies related to Vehicular Edge Computing have received a lot of attention from both academia and industry as an important part of Intelligent Traffic System and Internet of Vehicles.Unlike Mobile Edge Computing in the general sense,user movement in the Vehicular Edge Computing environment is characterized by faster speed and easy aggregation and dissolution in a short period of time,which exacerbates the uneven load of Vehicular Edge Computing servers.Therefore,it is more challenging to design load balancing methods for Vehicular Edge Computing.This thesis studies and analyzes the classical load balancing methods,summarizes the advantages and disadvantages of several types of load balancing methods,analyzes the characteristics and challenges of load balancing in Vehicular Edge Computing,and proposes an optimal control resource scheduling method based on load prediction in Vehicular Edge Computing(OCRS).The method establishes the state equation of the system by predicting the load of Vehicular Edge Computing servers,and then uses the equation to determine the load balancing policy and perform task migration in advance of the load balancing process using the predicted load values.This method is effective in balancing the load of each server,reducing latency and being anti-interference.The main work of this thesis is as follows.(1)A neural network-based edge load prediction method is proposed.This thesis first classifies the factors affecting the load into long-term and short-term factors,and then uses deep learning to predict the load.Since there are long-term factors and shortterm factors affecting the load,this thesis uses a multi-input neural network to deal with long-term factors and short-term factors separately.Finally,this thesis verifies the effectiveness of this edge load prediction method by platforms such as SUMO and Mob Fog Sim.(2)An edge resource load balancing method based on optimal control is proposed.In this thesis,the problem of solving the load balancing policy is described as an optimal control problem.The flow of the method is as follows: firstly,the parameters of the optimal control are generated based on the results of load prediction;then the load balancing policy function is solved based on the parameters,the input of which is the time and the load data of the on-board edge computing server;finally,the load balancing policy is solved.The edge resource load balancing method based on optimal control proposed in this thesis can set relevant parameters to adjust the load balancing effect,and its load balancing policy function has certain anti-interference capability.(3)The OCRS simulation system is implemented and verified using SUMO and Mob Fog Sim.In this thesis,OCRS and the same type of load balancing methods are compared and analyzed in different scenarios,and the results show that OCRS performs better in load balancing,delay and anti-interference.