Research And Development Of Driving Behavior Assessment And Driving Safety Management System Based On Edge Computing

With the continuous development of IOV technologies,people affect their attention on vehicle safety.The development of reasonable driving assessment methods has become an significant task to ensure safe driving.However,there are many assessment factors that affect driving safety,which makes the selection of driving indicators a major challenge.If driving indicators with too small influence are selected,it will lead to the lack of generality and objectivity.Nowadays,popular driving assessment applications mostly provide users with a general overview,which limits people get more quantitative information of driving assessment.In other words,users do not know at what time and due to what kind of behavior that lead to the decrease of their driving scores.To take full advantage of big data analysis,these driving assessment applications centralize computational tasks to cloud servers.With the rapid development of edge computing and microservices,cloud servers which previously highloaded and cohesive can effectively reduce their pressures and modularize highly coupled services.Therefore,developing a more representative method of driving behavior assessment and developing a less coupled assessment system become the core objectives of this paper’s research.This paper researches and improves existing methods of driving behavior assessment and their systems.Under the simulation environment combined with Air Sim and Unity.Through research and analysis,this paper selects behavioral indicators that directly affect driving safety.Then,it develops driving behavior assessment methods based on neural networks,which aims to improve the generality and objectivity of the driving assessment model.Finally,the assessment method is combined with edge computing,cloud services and clients to constitute a complete management system of driving safety.The main work of this paper includes the following three aspects.(1)We use the virtual OBD and virtual camera provided by Air Sim to collect the driving dataset from various drivers which consists of drivers’ behaviors and road images.Then,we use deep learning techniques to train neural networks of driving behaviors based on the dataset of skilled drivers.By studying different types of risk,comparing and clustering different styles of driving behaviors with the inferred output of the network,the range of values for different risk types is determined.(2)By establishing the model of driving behavior assessment,the weights of different driving behaviors in the simulation environment are determined.Then the corresponding scoring strategies integrated with inferred networks and risk types can finally lead to a complete method of driving behavior assessment.(3)This paper constructs a system of driving safety assessment based on the edge-cloudclient idea.The edge computing device is used to carry the driving behavior assessment method and calculate the real-time scoring of driving behavior;the cloud system is developed by using the microservice architecture to provide low coupling services for users;the Android APP is used to develop an interactive interface to provide users with query services of driving results.