Examining impacts of road landscapes on drivers' mental states, driving fatigue, and driving performance : a field experiment in the urban area

Abstract

The effects, if any, of the environmental features of urban roads on drivers’ mental states, driving fatigue and driving performance have rarely been examined via onsite driving experiments. To explore the influence of such features, my thesis involved a field study, in which 34 participants completed seven driving tasks on seven different routes in a randomized order at one-day intervals (except during bad weather). A total of 238 driving tasks were conducted, with each task lasted an hour. Participant’s mental states— including boredom, anger, frustration, anxiety, and avoidance— were measured immediately before, during, and after the driving task through visual analog scale questionnaires. Self-reported visual and muscular fatigue was also rated. BioHarness was used to measure heart rate continuously as a psychophysiological fatigue indicator. A Controller Area Network bus was used to measure driving performance continuously, including four parameters: mean of speed, standard deviation of speed, steering holds frequency, and steering reversal rate. Finally, participants reported their perceived environmental characteristics. Deep transfer learning for semantic segmentation was used to analyze road landscape characteristics and traffic conditions, which were recorded by the camera from the drivers’ view. Statistical analysis indicated three main findings on the relationships between environmental features and drivers’ mental states, fatigue, and performance, respectively. First, results of multi-factor ANOVA showed that roads with varying doses of greenness elicited significantly different mental states of drivers among seven routes after one-hour driving (F = 2.39, p < 0.05). The dose-response curve revealed a significant mental state promotion when the dose of greenness reached 39.21%. This mental state promotion effect of roadside greenness was greater for 60-minute driving than 30-minute driving. Second, I found that among environmental and traffic factors, road landscapes with high quantities and variations of greenness were the most effective in decreasing psychophysiological fatigue. A higher level of greenness was significantly associated with less psychophysiological fatigue (β = -3.24, p < 0.01), and more proportions of changes of greenness had a significant negative correlation with less psychophysiological fatigue (β = -2.77, p < 0.05). Third, I found positive effects of the proportion of roadside greenness on driving performance (β = 0.28, p < 0.001, adjusted R2 = 0.50). Mean proportion of roadside greenness contributes to as much as 20% of the model fit as measured by R2, suggesting that greenness is highly relevant to improving driving performance. Proportion of changes of greenness also has a significantly positive association with driving performance (β = 0.08, p < 0.05, adjusted R2 = 0.58). This consistent finding of the benefits of greenness adds to the knowledge of attention restoration theory and stress reduction theory by showing that natural benefits can still exist when driving. Another contribution of this study concerns the use of real-world driving experiments, which promises a high ecological validity of the study. Real-world driving experiments also provide opportunities to collecting real-time information on road environments and to using deep-learning methods for semantic segmentation analysis. This improves the understanding of detailed out-car environments and their impacts on urban driving activities.