Analysis And Prediction Methods For Driving Stress Loads And Traffic Accidents On Urban Roads Under The Influence Of A Combination Of Drivers,Vehicles,Roads And The Environment

With the rapid development of China's social economy and urbanisation,the prosperous development of urban traffic has also led to many social problems such as traffic safety.The core of an effective solution to urban road traffic safety problems is the scientific and reasonable analysis and prediction of traffic accidents.The urban traffic system is a dynamic coupled system consisting of four factors: driver,vehicle,road and environment.The safe driving process refers to the process of the driver manipulating the vehicle using safe and controlled driving behaviour under the influence of the combination of road and environment.Once the external stimuli of the combination of vehicle,road and environment exceed the driver's psychological expectations,it will lead to the driver adopting inappropriate driving behaviour and increase the risk of driving.The comfort and stability of the driver's psychological approach to driving ensures safe and controlled driving behaviour,and ensuring stable and appropriate driver psychological expectations is a fundamental way to improve traffic safety.However,most traditional traffic accident analysis and prediction methods consider external factors such as roads,vehicles and positioning analysis-based environment,ignoring the leading role of drivers themselves in safe driving,resulting in a lack of scientificity in accident analysis and prediction methods and reducing the effectiveness of traffic accident control.Road familiarity and stress load are used to describe driver factors.Road familiarity is an indicator that describes a driver's perception of the route,traffic environment and urban landscape,and leads to changes in their psychological condition through the influence of visual perception and feedback,which in turn leads to changes in driving behaviour and driving strategies.The impact of road familiarity on traffic safety is the result of a combination of driver,vehicle,road and environment factors.Traditional methods using only single factors such as driving behaviour and driving environment cannot reveal the mechanism of the impact of road familiarity on traffic safety well,thus missing key information and conclusions.And stress load refers to an indicator of the driver's need to reserve cognitive space after perceiving changes in vehicle,road and environmental factors during driving.An abnormal increase in stress load will cause the driving behaviour to fail to complete the expected operation,which will eventually act on the vehicle operating conditions and cause traffic accidents.However,stress monitoring and assessment methods are still difficult to apply on a daily basis due to the over-reliance on physiological data for quantifying stress load and the difficulty of collecting physiological data on a daily basis.Subsequently,by establishing a link with traffic safety,the mechanism of road familiarity and stress load in relation to driving safety is revealed,and based on this,a method for analysing and predicting urban traffic accidents under the influence of a combination of vehicle,road and environmental factors is constructed.In order to fill the gap between traditional accident analysis and prediction methods for driver factors,environmental factors and all-factor combinations,this paper establishes a reflection arc-like urban traffic accident analysis method based on the driver as the information interaction centre and the vehicle as the information transfer epiphenomenon from the perspective of the all-factor dynamic coupling system of driver,vehicle,road and environment in urban traffic system.Based on this,an all-factor data collection method and a multi-source data feature extraction and fusion method are proposed.Conducting real-world vehicle trials and using open source data such as rich,easily extractable and continuously growing street images as a data source for the full range of driver,vehicle,road and environment factors.Using multivariate techniques such as deep learning to extract real-time dynamic features of the full range of factors,and K-Means 3D cluster analysis to establish pressure load classification criteria,the project provides data support on vehicle,road and environmental factors for traffic accident analysis and prediction methods.The analysis and quantification of driver factors in terms of road familiarity and stress load.On the one hand,the impact of road familiarity on traffic safety is analysed in the context of driver physiological data,driving behaviour and the driving environment,and the mechanism of the impact of road familiarity on driver psychological conditions and even driving safety is investigated.On the other hand,the stress load is monitored dynamically and evaluated statically using all-factor real-time dynamic characteristics parameters.A machine learning algorithm is used to construct a driving stress monitoring model based on vehicle operation data and driving environment data,which outperforms most traditional algorithms without relying on physiological data.The model was constructed using open source data such as easily extracted and gradually improved street view images and machine learning algorithms,enabling large-scale and efficient assessment of pressure loads on urban roads,and also providing a data source for the construction of all-factor urban traffic accident prediction models.The interpretable Shapley Additive explained(SHAP)method was used to explore the importance of features to the model and the interaction between models,providing a quantitative tool to better understand the intrinsic link between the urban traffic environment and stress loads.The addition of urban pressure load assessment models,using only open source data such as street view images to extract the feature values of the full range of factors,constructed traffic accident prediction models for urban road sections based on the combination of driver,vehicle,road and environmental factors using Poisson regression models and negative binomial regression models,respectively,solves the problem of missing multi-source data collection and fusion methods,and promotes the process of transformation of urban road traffic accident prediction to efficiency and scale.The elasticity coefficient analysis method and marginal benefit coefficient analysis method explore the potential links between all-factor correlation variables and traffic safety.Finally,based on the results of the analysis and prediction of urban road pressure load and traffic accidents based on all factors of drivers,vehicles,roads and environment,a method for improving traffic safety based on all factors of drivers,vehicles,roads and environment is proposed,and a set of steps and process methods for improving urban traffic safety is established by combining the accurate prediction of accident-prone sections of urban roads with the pressure load analysis and the proposed improvement measures.This paper proposes a quantitative framework for the analysis and prediction of urban traffic accidents based on a combination of driver,vehicle,road and environment factors,and introduces road familiarity and stress load indicators to quantify the influence mechanism of driver factors on traffic safety.The paper also provides a theoretical supplement to the traditional urban traffic accident analysis and prediction methods from the perspective of all factors with the driver factor as the core,and achieves accurate prediction of urban traffic accidents with all factors,and provides theoretical support and methodological support for the improvement of urban road traffic safety.