Analysis And Modeling Of Driver's Mental Workload Under Low Illumination At Dusk

According to traffic accident statistic data,dusk is the peak time of accidents in a day.The reason is that the rapid decline of illumination affects drivers' perceptions of surrounding traffic environment,then their mental workload increase accordingly.In addition,it brings forward higher request on how to react and control on complex urban road conditions in the evening peak hours,which may result in the lack of alertness,operational errors and other dangerous behaviors.Therefore,the research on the mental workload of drivers in the condition of low illumination at dusk,has great theoretical and practical significance to ensure the safety of urban driving and to reduce the traffic accident rate.Involved with traffic engineering,human factors engineering,statistics,artificial intelligence and some other subjects,this paper considered the interaction relationship of the elements "human-car-road-environment",collected multi-source data of drivers and vehicles under low illumination at dusk in urban road through real-vehicle experiment.And this article established the index set of mental workload,and studied driver's mental workload prediction model in depth,on the basis of analyzing drivers' physiological and behavioral characteristics and the relevant function.Firstly,started from the concept of driver's mental workload,drivers' physiological and behavior characteristics under low illumination at dusk were analyzed,and mental workload indices were selected preliminarily from multi-dimension.Afterwards,the paper designed the vehicle experiment scheme for the low illumination environment,and the environmental illumination level was divided scientifically,and then we collected physiological and behavioral characteristics of drivers,while subjective mental workload index was acquired by NASA-Task Load Index.And then,this paper used statistical methods to analyze the influence of illumination and driving proficiency on the physiological and behavioral characteristics as well as their correlation,then we attempted to establish a trend fitting regression model between environmental illumination and key indicators.The results revealed that the drop of environmental illumination inceased drivers' mental workload,and decreased their driving stability as well,while there were significant differences between novice drivers and skilled drivers.Finally,this paper established a mental workload characterization index set using the Correlation Analysis and Principal Component Analysis,and constructed driver's mental workload model under the condition of low illumination at dusk base on kNN,SVM and GBDT.The effects of model were evaluated with the help of measured data,while the effectiveness of the index set was evaluated by eliminating individual indicators.The results showed that the index set was representative,and the drivers' mental workload model under the condition of low illumination at dusk based on GBDT algorithm,had acquired good recognition results reaching to 92.25%.The research result will guide the driver to drive safely,scientifically and reasonably,and provide the basis for traffic active safety management and the design of the vehicle warning terminal system under the background of "intelligent traffic".