Research On The Mechanism And Prediction Model Of Driving Mental Workload Based On Attention Deman

Road traffic accidents directly caused by drivers’ human fault accounted for 70.5%.The research of drivers’ mental workload(DMW)accompanied with driving maneuvers plays an important role in the field of controlling human fault in traffic accidents.On the other hand,the multidimensional measurements of DMW,the correlation and interaction of its influencing factors,which made the difficulty for researching the mechanics of DMW.Therefore,the research of the internal DMW by the external road traffic environments and driving behavior’s performance based on the systematic point of view,which will contribute to reveal and understand the process and rules of DMW,and then provide a theoretical basis for research of human factors in traffic safety.A structural DMW model was built based on the resource supply theory.Combined with its relationship of mental capacity,information processing and performance,the practicability of DMW represented by the visual attention demand was explored.From the point of view of the attention load margin,the DML was defined again;starting from the amount of information about the road and traffic accepted by drivers.Two non-dimensional quantitative indicators:the consistency of linear combination and the reciprocal of headway-as the measure of road geometry alignment and traffic complexity were proposed respectively;Through out the analysis of DMW’s process,a structural DMW model was built.The d structural model explained the influencing and reciprocal relationships of the DMW,which indicated that the DMW was not only influenced by the difficulty of the driving task,but also by driving performance feedback.In accordance with the methodology of the system,a conceptual model of evaluating and predicting the DMW was constructed.The researches on influencing and affecting mechanism of DMW were carried out by adopting the method of driving simulation experimental study.Firstly,two virtual experiment fields with typical road geometries and traffic flows were constructed.Driving simulation experiments of 40 typical road sections were performed in the virtual environment by 23 skilled drivers selected.The "Kuming Road Traffic Driving Simulation" and "Attention Demand Test System" were used to record the driving behavior,vehicle’s responding performance(deriving behavior performance)and attention demand rate of the driver subjects.Afterwards,based on simulation experiments,the interrelationships amony the road traffic complexity,driver behavior performances and DMWs were analyzed,and their interrelationship rules were revealed.On the road geometry experiment conditions,in which the driver subjects should control speed according to requests,the geometric complexity was positively correlated with the lateral lane deviation,lateral speed,lateral acceleration,steering angle and braking behavior,whereas it negatively correlated to longitudinal speed,longitudinal acceleration and gear.There is a significant quadratic function relationship among the driver workload characterized by attention demands and the geometric complexity.Among driving behavior parameters,the mean value,standard deviation and maximum value of lateral lane deviation,longitudinal speed,lateral speed,acceleration and steering angle are positively related to the driver workload characterized by attention demands;while accelerator,brake and gear are related to the driver’s driving strategy.Under the road traffic flow experiment conditions,the complexity of traffic flow and driver’s individual driving strategy significantly affect driver behavior,lateral lane deviation,longitudinal acceleration,steering angle and braking are positively correlated to traffic complexity,while longitudinal speed,lateral speed,lateral acceleration,accelerator,gear are negatively correlated to traffic complexity.the changes of traffic flow significantly affect driver workload,according to the order of "free flow-steady flow-unstable flow-forced flow",the attention demand correspondingly presents "high-low-second high-high" similar to "U" type trend.Among the indicators of driving behavior parameters,attention demand has a strong negative correlation with the mean value and standard deviation of lateral acceleration;while has a positive correlation with the standard deviation of lateral lane deviation and the average of accelerator pedal.The dynamic modulatory strategy of DMW were discovered and its universal rule was proposed:drivers are inclined to keep a certain desired level of DMW(the objective DMW),and accommodate it by their speeds.The methods of regression analysis and curve fitting were applied to analyze the relationship among DMW characterized by attention demands,complexity and speed.The analysis results showed that DMW was not only dependent on the geometric and road traffic complexity,but also the speed strategy.The dynamic modulatory strategy of DMW presented obviously in the free flow and steady flow,which with low traffic complexity.Then another experiment was taken to explore the quantitative relation between the speed and DMW.It proved that a linear correlation between them existed.A prediction model and method for quantifying DMW were constructed.Using principal component analysis,data mining to the driver behavior parameters based on DMW were carried on.Nine driving behavior factors such as lateral behavior,longitudinal speed,lateral speed and lateral lane deviation adjustment etc.were proposed to reveal the potential driving behavior characteristics,and largely eliminated the multicollinearity among driver behavior parameters.The BP and RBF neural network algorithm of artificial neural network method were adopted to build the dynamic prediction model of driver workload with the nine factors as input layer.A feasible method for evaluateing DMW was put forward.Among the two methods,the training accuracy of the prediction model of BP algorithm was 89.374%and 83.089%respectively;while the training accuracy of the prediction model of RBF algorithm was 87.593%and 79.784%respectively.The model comparison of predicting results shows the RBF method was slightly better.Finally,the actual road experiment was taken to verify the attention demand test results and driver behavior respectively in the virtual environment and real environment,and the conclusion proved the validity of the model.