The Research About Drive Distraction Based On Driver-in-loop Simulation Experiments

With the development of new in-vehicle technology,drivers are exposed to more sources of distraction,which makes driver distraction a major cause of unsafe driving.Compared with fatigue,distraction features have shorter duration,unlike fatigue driver can recover swiftly from distraction.Influencing factor of distraction are sophisticated,different type of distraction have diverse effect,driving contexts also influence driving performance.To solve those problems above,this paper conduct a research on driver distraction based on eye movement and driving performance.The whole experiments were based on driving simulation,a typical driving scenarios were reconstructed;There are two types of distraction in the paper:visual distraction and cognitive distraction.The statistical analysis of candidate features from eye movement and driving performance was applied first.The optimal feature subsets were extracted by using Relief algorithm.Finally based on the GMM statistical model,the extracted features as input,a distraction detection algorithm was designed and cross validated,the influence caused by secondary tasks and driving contexts are discussed separately.The main research work of this study are follows:(1)Typical driving scenarios were built,arrow direction recognition and auditory continuous memory test were introduced as visual and cognitive secondary tasks to induce driver into distraction driving.Eye movement and driving data were collected under two condition:normal driving and distracted driving.Ten subjects participated in the experiment,and outliers were removed.(2)Student’s T test was conducted to examine the level of saliency.The features of eye movement and driving performance under different condition were summarized and statistically analyzed.The results indicate the these features were impacted by the distracted type and driving contexts.According risk management theory from cognitive psychology,explain certain overcompensation and overcompensation behavior.Generally speaking the visual task have greater influence on driving performance,driver also more sensitive to visual distraction.(3)Total 27 feature were extracted from time sequence information of steering operation and vehicle state parameters speed.To cope with high-dimension and small sample size of collected data,Relief algorithm a is adopted to extract the optimal feature subset for best classification performance.This algorithm calculate weight of all features,top 5 features were selected as optimal feature subset.The optimal feature subset for visual secondary task and cognitive secondary task was selected separately.A driver distraction detection algorithm based on GMM was designed and validated.This algorithm use driving performance features as inputs.The algorithm gets correct rates on average between 64.4%-82.6%.Effect of both secondary task and driving context was analyzed,through this discussion,the Yerkes-Dodson Law is validated and the importance of driving context is proved.