Study On Effects Of Traffic Environment And Driving Experience On Driver's Eye Movement And Workload

Among all reasons that induced road traffic accidents, about 90% are related to human factors, in which the perception factor plays the primary role. It is deemed that 90% of information perception of car drivers comes from vision. It is obvious that vision is of great importance in driving task and accident control. The study on the basic visual behavior of eye movement will help to discover and understand the process and laws in driver's object search and potential risk perception. The study also has important significance of theoretical instruction for improving the design guidelines of roadway and vehicles, for amending driver's training and testing rules, and for designing new-type driving assistant system and risk early warning system.The driving tests were carried out in 4 typical traffic environments, including urban roadway, normal highway, mountainous highway and suburban highway. Totally 24 object drivers with different driving experiences were chosen. In driving process, the eye movement data such as the fixation duration, fixation object, saccade duration and saccade velocity were recorded by high speed eye movement tracking system (eye movement record device). The physiological signals of driver's heart electrical activity, galvanic skin response (GSR) and breath frequency were recorded by physiological test device. The GPS was used to record vehicle speed in real time during the whole test process.The test data were analyzed with statistical method. Firstly, the variational rules of fixation duration, visual search variance, saccade velocity and saccade amplitude of drivers with road conditions were investigated statistically. Secondly, by counting the fixation objects one by one, drivers' fixation duration and fixation frequencies at each type of object were analyzed, and areas of fixations (AOFs) were divided according to object positions. On that basis, the variances between experienced drivers and inexperienced drivers in different traffic environments were explored focused on fixation frequencies at each type of object and fixation distribution in different AOFs. Drivers' visual field was divided into 6 AOFs by dynamic cluster analysis. Using Markov chain theory, driver's fixation transition mode was studied, in which the one-step transition probabilities among drivers' different AOFs were solved and the Markov stationary distributions in each AOF were obtained. In addition, driver's workload was characterized by fluctuating rate of heart rate and heart rate variability. The effects of traffic environments and driving experiences on driver's workload were explored, and the correlation analysis of eye movement and workload was carried out. Lastly, 8 indexes of fixation transition modes and 5 composite indexes were chosen, and two evaluation models on driving experience were built by using principal component analysis (PCA), the effects of each index on driving experience were investigated as well.Study results indicate that, traffic environment and driving experience have significant effects on driver's eye movement behavior and workload. The following conclusions can be drawn from this research work.1. Driver's fixation duration, horizontal and vertical variances of visual search, saccade amplitude and saccade velocity vary with change of road conditions. Drivers adjust their fixation object and distribution of attention according to the traffic environment.2. Compared to inexperienced drivers, experienced drivers show a more flexible visual search mode.3. Inexperience drivers endure heavier workload than experienced drivers in the same traffic environment.4. Driver's mental effort and workload levels can be reflected by the saccade behavior.5. The 6 AOFs for driver's visual field divided by the method of dynamic clustering are in accord with the actual AOFs distribution.6. Inexperienced drivers can be efficiently distinguished from experienced ones with the evaluation models built by PCA method.The research was sponsored by National Natural Science Foundation (50678027).