The extent and characteristics of driving anxiety

Research over the last two decades has resulted in improved understandings of the nature and characteristics of driving anxiety. However, we still do not know the extent of driving anxiety in the general population, as most studies have focused on clinical or vehicle accident samples, and the only population study is of older adults. The present study addressed this gap in knowledge using data from 441 people who responded to a survey sent to a random sample of 1500 adults recruited from the compulsory voting register in New Zealand. While 31% reported no driving anxiety, 52% endorsed mild driving anxiety and 16% reported moderate to severe driving anxiety. There were small but significant differences in the average age at which participants in these three groups started to learn to drive, but there were no differences in driving-related accidents and incidents over the past 12 months. Those with higher driving anxiety endorsed more anxiety about road rage, lower feelings of safety while driving, and more driving-related avoidance and negative cognitions than the less anxious participants. However, levels of helpseeking were low, and suggest the need to develop online self-help packages that are more accessible and acceptable to those experiencing driving anxiety.     

Modelling the human lane-change execution behaviour through Multilayer Perceptrons and Convolutional Neural Networks

Driving is a highly complex task that involves the execution of multiple cognitive tasks belonging to different levels of abstraction. Traffic emerges from the interaction of a big number of agents implementing those behaviours, but until recent years, modelling it by the interaction of these agents in the so called micro-simulators was a nearly impossible task as their number grows. However, with the growing computing power it is possible to model increasingly large quantities of individual vehicles according to their individual behaviours. These models are usually composed of two sub-models for two well-defined tasks: car-following and lane-change. In the case of lane-change the literature proposes many different models, but few of them use Computational Intelligence (CI) techniques, and much less use personalization for reaching individual granularity. This study explores one of the two aspects of the lane-change called lane-change acceptance, where the driver performs or not a lane-change given his intention and the vehicle environment. We demonstrate how the lane-change acceptance of a specific driver can be learned from his lane change intention and surrounding environment in an urban scenario using CI techniques such as feed-forward Artificial Neural Network (ANN). We work with Multilayer Perceptron (MLP) and Convolutional Neural Networks (CNN) architectures. How they perform one against the other and how the different topologies affect both to the generalization of the problem and the learning process are studied.     

Driving anger and its relationship with aggressive driving among Chinese drivers

Driving anger poses a serious threat to road safety. Increasing attention is being paid on this issue, with driving anger usually measured by a 14-item version of the Driving Anger Scale (short DAS). However, driving anger problem in China has received limited research attention and there is no corresponding Chinese version of the short DAS. This study adapted the short DAS for use with Chinese drivers and investigated the relationship between driving anger and aggressive driving with an Internet-based survey conducted to a sample of Chinese drivers. The Confirmatory Factor Analysis results showed that a three-factor DAS structure provided a good fit to the data obtained, with the three subscales used being hostile gesture, safety-blocking and arrival-blocking. The hostile gesture subscale and arrival-blocking subscale were positive predictors while the safety-blocking subscale was a negative predictor of aggressive driving. In China, the overall driving anger was lower but its association with aggressive driving was stronger, than that in western countries. These findings provide important insights into causes and consequences of driving anger for the development of effective strategies to reduce driving anger and to enhance road safety.     

Predicting intentions to text and call while driving using the theory of planned behaviour

There is extensive evidence that using a mobile phone while driving causes degradation in driving performance, and thereby results in reduced safety on the road. The present study examined intentions to use mobile phones while driving using the Theory of Planned Behaviour (TPB). A total of 212 Ukrainian drivers (mean age = 35 years SD = 10 years; males = 82%) completed a survey that included measures of the TPB components related to intentions to send or read text messages or to make or receive handheld phone calls across two different scenarios; one where they were running late, and the other when they were not in a hurry. Measures of the frequency of mobile phone use were also collected. The results showed that 63% of the sample reported using a mobile phone while driving at least daily, with the most frequent types of usage being making and answering a phone call with a handheld device. The most consistent predictor of intentions to interact with a mobile phone while driving was having a positive attitude towards doing so. Perceived behavioural control was also significantly and positively associated with mobile phone use while driving, but only a small number of associations were found with subjective norms. Our results suggest that intentions to interact with mobile phones while driving may be context specific.     

Attention during visual secondary tasks in driving: Adaptation to the demands of the driving task

Visual attention in driving with visual secondary task is compared for two visual secondary tasks. N = 40 subjects completed a 1 h test drive in a motion-base driving simulator. During the drive, participants either solved an externally paced, highly demanding visual task or a self paced menu system task. The secondary tasks were offered in defined critical and non-critical driving situations. Eye movement behavior was analyzed and compared for both tasks. Before starting the secondary tasks, eye movement behavior shows a smaller standard deviation of gaze as well as longer fixation durations for both tasks. The comparison between the two tasks indicates that drivers use the possibilities the self paced task offers: during the secondary task, they monitor the driving scene with longer fixations and show a greater standard deviation of gaze position. Furthermore, independently of the type of secondary task, drivers adapt their eye movement behavior to the demands of the situation. In critical driving situations they direct a larger proportion of glance time to the driving task. Last, the relation between glance behavior and collisions is analyzed. Results indicate that collisions go together with an inadequate distribution of attention during distraction. The results are interpreted regarding the attentional processes involved in driving with visual secondary tasks. Based on the similarities and differences between the two secondary tasks, a cognitive approach is developed which assumes that the control of attention during distraction is based on a mental situational model of the driving situation.     

Driving simulator validation with hazard perception

How should we assess the comparability of driving on a road and “driving” in a simulator? If similar patterns of behaviour are observed, with similar differences between individuals, then we can conclude that driving in the simulator will deliver representative results and the advantages of simulators (controlled environments, hazardous situations) can be appreciated. To evaluate a driving simulator here we compare hazard detection while driving on roads, while watching short film clips recorded from a vehicle moving through traffic, and while driving through a simulated city in a fully instrumented fixed-base simulator with a 90 degree forward view (plus mirrors) that is under the speed/direction control of the driver. In all three situations we find increased scanning by more experienced and especially professional drivers, and earlier eye fixations on hazardous objects for experienced drivers. This comparability encourages the use of simulators in drivers training and testing.     

Age and gender comparisons of driving while sleepy: Behaviours and risk perceptions

To assist road safety professionals in developing effective strategies to combat the risk associated with driving while fatigued, a survey was administered to 1000 Australian drivers. Participants reported their past behaviours in regards to driving while sleepy and their perceptions of risk associated with driving fatigued as compared to speeding and driving under the influence of alcohol. Although participants appeared to be aware of the substantial risk associated with driving while sleepy, many drivers reported that they frequently drive when sleepy. Age and gender comparisons, revealed that risk taking behaviour in regards to driving while sleepy is occurring across all age groups and in both male and female drivers. Overall young to middle age drivers and male drivers reported the highest frequency of driving while sleepy and reported the lowest perceived personal risk in regards to driving while sleepy.     

Do complex models increase prediction of complex behaviours? Predicting driving ability in people with brain disorders

Prediction of complex behavioural tasks via relatively simple modelling techniques, such as logistic regression and discriminant analysis, often has limited success. We hypothesized that to more accurately model complex behaviour, more complex models, such as kernel-based methods, would be needed. To test this hypothesis, we assessed the value of six modelling approaches for predicting driving ability based on performance on computerized sensory–motor and cognitive tests (SMCTests?) in 501 people with brain disorders. The models included three models previously used to predict driving ability (discriminant analysis, DA; binary logistic regression, BLR; and nonlinear causal resource analysis, NCRA) and three kernel methods (support vector machine, SVM; product kernel density, PK; and kernel product density, KP). At the classification level, two kernel methods were substantially more accurate at classifying on-road pass or fail (SVM 99.6%, PK 99.8%) than the other models (DA 76%, BLR 78%, NCRA 74%, KP 81%). However, accuracy decreased substantially for all of the kernel models when cross-validation techniques were used to estimate prediction of on-road pass or fail in an independent referral group (SVM 73–76%, PK 72–73%, KP 71–72%) but decreased only slightly for DA (74–75%) and BLR (75–76%). Cross-validation of NCRA was not possible. In conclusion, while kernel-based models are successful at modelling complex data at a classification level, this is likely to be due to overfitting of the data, which does not lead to an improvement in accuracy in independent data over and above the accuracy of other less complex modelling techniques.     

Tactile detection response task: Metrics for assessing drivers’ cognitive workload

The Tactile Detection Response Task (TDRT) has been used to assess the cognitive workload of driver distraction with response time and miss rate as metrics of cognitive workload. However, it is not clear which metric is more sensitive and whether sensitivity is maintained for visual tasks. The objective of this study was to assess the sensitivity of the TDRT to changes in cognitive workload and to examine whether the sensitivity depends on task modality. A driving simulator study was conducted with 24 participants. The study included restaurant selection tasks with three presentation modalities (auditory, visual, and hybrid) and two difficulty levels (low and high). The high difficulty level was designed to be more cognitively demanding than the low difficulty level. Mixed-effects models were applied to examine the TDRT metrics and task difficulty level. The model controlled for age group, gender, and included a random effect for participants. The high difficulty level of the auditory tasks significantly increased the likelihood of missing a TDRT stimulus. No statistically significant differences were observed for visual and hybrid tasks. TDRT response time was not significantly associated with the difficulty level, regardless of task modality. In this study, the binary outcome TDRT miss was thus considered a more sensitive metric of cognitive workload than TDRT response time. TDRT response time can still be used to measure cognitive workload when tasks are relatively easy and the TDRT miss rate is close to zero. In addition, the sensitivity of the TDRT miss diminished for tasks that involved a visual component. Researchers who use TDRT to measure the cognitive workload associated with visual tasks should be aware of this limitation.