Research On Safety Prediction Technology Based On Edge Intelligence In Vehicle Communication

The rapid development of IoT has lead to the rapid development of the service-driven intelligent network.Meanwhile,vehicular communication has attracted much attention as an important application of intelligent transportation systems.However,the current task scheduling algorithms used in the edge computing environment of VANET cannot meet the requirements of millisecond delay and low fault tolerance for services such as security prediction and intelligent traffic control.Therefore,this paper focuses on the edge computing environment in vehicular communication to study the driving safety prediction technology based on edge deep learning for vehicle self-organizing network.By proposing a new deep learning task scheduling algorithm and a new driving safety prediction algorithm,the delay of driving safety prediction is reduced and the efficiency and accuracy of driving safety prediction are improved,the main research work of this paper is as follows:(1)Aiming at the problem that the task scheduling algorithm used in the current edge computing environment generates too high latency and low efficiency when scheduling deep learning tasks,this paper proposes a task scheduling algorithm in the edge deep learning environment of vehicle self-organizing network.The algorithm firstly offloads part of the driving safety prediction tasks from the centralized cloud server to the edge servers near the vehicle networking equipment,and determines the number of layers of the neural network that the cloud server offloads to the edge server according to the calculation overhead and the bandwidth requirement generated by each deep learning task,thereby more effectively utilizing the computing resources of the edge server.The simulation results show that compared with the task scheduling algorithm,the task scheduling algorithm effectively reduces the data size transmitted from the edge server to the cloud server,releases the bandwidth resources of the network and reduces the transmission delay.The scheduling algorithm proposed in this paper can improve the driving safety prediction efficiency under the condition of satisfying delay.(2)Aiming at the problems of low prediction accuracy caused by the feature interactions between the driving information,the road condition information and the driver information,and the problem of the low prediction efficiency caused by a large number of feature engineering,a factorization machine combined neural network(FMCNN)is proposed to predict the driving safety of vehicles in VANET.In the proposed framework,the factorization machine and the deep neural network are used to learn the effects of low–order and high-order feature interactions from the driving information and the weather information in the pre-training phase respectively,thereby improve the learning ability of deep neural network.After the pre-training phase,the high-order feature interactions extracted by the last hidden layer of the deep neural network and the low–order feature interactions trained by the factorization machine are the input of a new deep neural network.Finally,the vehicle's driving safety is predicted based on the output of the neural network.The experiment results show that the prediction result of the proposed FMCNN is better than DNN and FM in AUC and Logloss.