Behavior Spectrum-Based Tunnel Driving Behavior Analysis And Self-Explaining Design

The special spatio-temporal environment of tunnels,with their small space,monotonous environment,high noise level,and large differences between internal and external landscapes,is prone to adversely affect drivers,and the severity and cost of traffic accidents on tunnel sections are generally higher than those on ordinary sections.Therefore,under the multi-factor coupling of human-vehicle-road-environment,it is important to study the driving behavior of drivers in each section of the tunnel and assist them to complete the safe transition between the tunnel section and the normal section as well as between the sections of the tunnel,which is important to improve the level of tunnel traffic safety.Based on the above considerations,this paper,based on the natural driving test,starts the research from quantifying the complexity of tunnel operation environment,constructing multi-dimensional characteristic driving behavior spectrum and tunnel self-explanatory road design.First,based on the real vehicle test,this paper uses CAN-OBD analyzer,heart rate meter and eye movement meter to obtain driving behavior,psychological and physiological data of30 subjects in the long tunnel(cluster)section,pre-process the data and extract 7characteristic indicators describing driving status.The data were pre-processed and 7characteristics describing the driving state were extracted.The multi-dimensional characteristics of the operating environment such as sound,light,number of vehicles and tunnel length were extracted from the test video,and the differences in the environmental characteristics of each section of the long tunnel and the tunnel group were compared using mathematical and statistical methods,and then the effects of the environmental characteristics of different sections on driver behavior,psychology and physiology were studied by correlation analysis.At the same time,a fuzzy comprehensive evaluation method was used to classify the complexity of the tunnel operating environment into three levels: low,medium and high,and to quantify the comprehensive environmental score.Then,recurrence diagram and spectral radius are introduced to propose single feature recurrence matrix spectral radius for the first time,extract the recurrence matrix spectral radius index of multidimensional features in time series,calculate the mean value of the index and the change rate of the mean value,and analyze the change law of driver behavior,psychology,physiology and operating environment features with road sections and zones.Based on the recurrence matrix spectrum radius index,the driving behavior spectrum is constructed,and the driving state of the behavior spectrum is classified into integrated smooth and integrated non-smooth by the Hidden Markov Model and CRITIC(Criteria Importance Though Intercriteria Correlation)weighting method,so as to determine the time or section where the driver’s state fluctuates.This method is used to determine the time or zone when the driver’s state fluctuates.At the same time,a comprehensive driving state classification model based on Light Gradient Boosting Machine(Light GBM)is constructed to verify the effectiveness of the multidimensional feature composite recursive matrix spectrum radius indicator in identifying comprehensive driving states.Finally,based on the driving behavior spectrum,the behavioral,psychological and physiological changes of drivers in each segment are analyzed.On this basis,the long tunnels and tunnel groups are divided into variable self-explanatory segments,transitional self-explanatory segments and constant self-explanatory segments along the longitudinal spatial dimension,and the self-explanatory characteristics of the three segments are distinguished by combining the spectral features to give design countermeasures in terms of pavement marking,colored pavement,roadside landscape and tunnel cavity design.And then design a group of long tunnel self-interpretation road scheme,and verify the effectiveness of self-interpretation road through driving simulation test.The research results show that: the extra-long tunnels operate with the highest environmental complexity,and the distribution of environmental complexity is more similar between the tunnel group sections and the normal sections.The recurrence matrix spectral radius can effectively characterize the fluctuation of the data compared to the standard deviation.The multidimensional feature behavior spectrum constructed based on the recurrence matrix spectrum radius can visually reproduce the changes of drivers’ operating environment,driving behavior,psychological and physiological changes in the tunnel sections.The multidimensional feature composite recursive matrix spectrum radius can effectively identify the comprehensive driving state with a model accuracy of 81.1%,and the highest accuracy can reach 84% after debugging the model parameters.It was found that drivers are prone to driving state fluctuations in the entrance section of extra-long tunnels,the transition section of extra-long tunnels and the connecting section of tunnel groups,and the driving state is most stable in the middle section of extra-long tunnels.When designing self-explanatory roads for tunnels,the variable section needs to reduce the driver’s speed,relieve the psychological load and assist the driver to complete the spatio-temporal transition;the transition section needs to relieve the driver’s tension,assist the driver to quickly adapt to the dark environment and avoid rapid acceleration;the unchanged section needs to improve the driver’s attention,relieve driving fatigue and enhance the driver’s perception of speed.The results of the study can provide technical support for analyzing the spatio-temporal characteristics of tunnel driving behavior and improving traffic safety.