Improving Fitness to Drive: The Case for Hazard Perception Training

Predicting collision cases at unsignalized intersections using EEG metrics and driving simulator platform

Unsignalized intersection collision has been one of the most dangerous accidents in the world. How to identify road hazards and predict the potential intersection collision ahead are challenging problems in traffic safety. This paper studies the feasibility of EEG metrics to forecast road hazards and presents an improved neural network model to predict intersection collision based on EEG metrics and driving behavior. It is demonstrated that EEG metrics show significant differences between collision and non-collision cases. It indicates that EEG metrics can serve as effective indicators to predict the collision probability. The drivers with higher relative power in fast frequency band (alpha and beta), lower relative power in slow frequency band (delta and theta) are more likely to have conflicts. The prediction using three machine learning models (Multi-layer perceptron (MLP), Logistic regression (LR) and Random forest (RF)) based on three input datasets (only EEG metrics, only driving behavior and combined EEG metrics with driving behavior) are compared. The results show that for single time point prediction, MLP model has the highest accuracy among three machine learning models. The model solely based on EEG metrics datasets has higher accuracy than driving behavior as well as combined datasets. However, for multi-time point prediction, the accuracy of MLP is only 73.9%, worse than LR and RF. We improved the MLP model by adding attention mechanism layer and using random forest model to select important features. As a consequence, the accuracy is greatly improved and reaches 88%. This study demonstrates the importance and feasibility of EEG signals to identify unsafe drivers ahead. The improved neural network model can be helpful to reduce intersection accidents and improve traffic safety.

Improved Perception of Motorcycles by Simulator-Based Driving Education

Research shows that about half of all motorcycle collisions with other vehicles were caused by the accident opponent, typically a passenger car. This study aimed to assess the effect of simulator training on improving car drivers’ perceptibility of motorcycles and thereby addressing this frequent type of motorcycle accident from the perspective of the initiator. For this purpose, a training program with different methods was conducted and tested in a driving simulator with 80 learner drivers aged between 15 and 27 years, assigned to a control group and three training groups: variable priority, equal priority, and equal priority with warning. The conflict scenarios were determined based on an analysis of motorcycle–car accidents. The variable priority training program resulted in better perceptibility of motorcycles as compared to the equal priority training program and equal priority with warning in two out of four test setups, i.e., urban roads with high contrast between motorcycle and the driving environment and on rural roads with a low contrast. Most participants rated each training method in the driving simulator as useful and would recommend it to other learner drivers. These results are important because they show that simulator training has a positive effect on the motorcycle detection performance of learner drivers. The early perception of motorcycles in car drivers is essential for preventing collisions between cars and motorcycles.

Identifying objective behavioural measures of expert driver situation awareness

Efforts to reduce road crash rates depend on a clear understanding of the factors that contribute to driver risk. Not all drivers are at equivalent risk. It is critical to identify the factors that contribute to the development of expertise in the driving environment. The rapid development of a driver's situation awareness (SA) is central to the safe performance of the driving task. Therefore, SA must be clearly operationalised in order to better assess its role in the development of expertise. This study employs an existing scheme based on the Perceptual Cycle Model (PCM) used for post hoc incident analysis and adapts it to the driving context. We attempted to correlate performance on coded verbalisations indicative of SA with non-invasive objective gaze metrics. Gaze metrics and the verbal counts were shown to differentiate between both expert and experienced (non-expert) drivers, but these measures failed to correlate with one another. Findings indicate differences in the way expert and experienced drivers update their schema of the driving task, with equivalent effort required to do so. The novel adaptation demonstrated in this paper allows for a domain-specific assessment of SA which reliably differentiates between drivers of varying expertise levels. Although selected gaze metrics were shown to be inadequate predictors of SA, additional analysis demonstrated key differences in gaze content. Combined, these findings enhance an understanding of expert SA development contributing to reduced crash risk.

Learner drivers (and their parent-supervisors) benefit from an online hazard perception course incorporating evidence-based training strategies and extensive crash footage

We previously found that a six-session online hazard perception training course, which incorporates evidence-based learning strategies and footage of over a hundred real crashes, improved hazard perception skill and reduced risk-taking intentions in novice drivers who had passed their on-road driving test within the previous three years. However, one issue with targeting crash-prevention training at individuals who are already driving unsupervised is that drivers are at their highest crash risk immediately after they pass their on-road driving test. That is, the training may arrive too late to protect drivers while they are at their most vulnerable. It is also possible that it may prove difficult to persuade drivers to complete an unsupervised training course if they are already licensed to drive independently. Given that learner drivers cannot drive unsupervised, and that they are typically supervised by a parent, one potential strategy is to target the training at learners and to ask their parents to provide one-on-one mentoring throughout the course. We therefore recruited learner driver/parent-supervisor dyads to participate in a randomized control study, with the objective of examining the effects of the hazard perception training course on aspects of driving behaviour associated with crash risk (as measured using validated computer-based tests). Outcome measures included two hazard perception skill assessments (a response time hazard perception test and a verbal response hazard prediction test), and three tests assessing aspects of risk-taking propensity in driving (speed choice, following distance, and gap acceptance). Learners who completed the course (N = 26) significantly improved their scores on both hazard perception skill measures, and also chose safer following distances, compared with a waitlist control group (N = 23). However, the training did not significantly reduce learners' speed choice or gap acceptance propensity. The hazard perception skill of parent-supervisors, who observed the course but did not complete it, also improved on both hazard perception measures, relative to controls. Additionally, both learners and their parent-supervisors reported a range of positive effects on the learners' real-world driving performance. These results suggest that this type of hazard perception training could be beneficial if deployed during the learner phase of driver licensing.

Identifying Interactive Factors That May Increase Crash Risk between Young Drivers and Trucks: A Narrative Review

Sharing the road with trucks is associated with increased risk of serious injury and death for passenger vehicle drivers. However, the onus for minimising risk lies not just with truck drivers; other drivers must understand the unique performance limitations of trucks associated with stopping distances, blind spots, and turning manoeuverability, so they can suitably act and react around trucks. Given the paucity of research aimed at understanding the specific crash risk vulnerability of young drivers around trucks, the authors employ a narrative review methodology that brings together evidence from both truck and young driver road safety research domains, as well as data regarding known crash risks for each driving cohort, to gain a comprehensive understanding of what young drivers are likely to know about heavy vehicle performance limitations, where there may be gaps in their understanding, and how this could potentially increase crash risk. We then review literature regarding the human factors affecting young drivers to understand how perceptual immaturity and engagement in risky driving behaviours are likely to compound risk regarding both the frequency and severity of collision between trucks and young drivers. Finally, we review current targeted educational initiatives and suggest that simply raising awareness of truck limitations is insufficient. We propose that further research is needed to ensure initiatives aimed at increasing young driver awareness of trucks and truck safety are evidence-based, undergo rigorous evaluation, and are delivered in a way that aims to (i) increase young driver risk perception skills, and (ii) reduce risky driving behaviour around trucks.

A comparison of motorcycle instructor candidate selection practices in the United States

INTRODUCTION

An essential aspect of motorcycle rider education is how the instructor selection process impacts student learning, sometimes referred to as the human element, as it is a significant factor influencing curriculum success. Student and program achievements are partially contingent on instructors who understand the curriculum and facilitate student learning during instruction. Previous research on motorcycle rider education has emphasized a need for the examination of instructor selection and development, stating that quality education is reliant on instructors who are competent and qualified.

METHOD

By applying an exploratory study method, state and military Motorcycle Safety Education Program Managers and Instructor Trainers were examined and compared through telephonic interviews to develop a greater understanding of instructor candidate selection criteria and vetting processes.

RESULTS

The results suggest that changes in instructor candidate selection systems may improve decisions about a candidate's job and organizational fit.

CONCLUSIONS

Study conclusions indicate that use of multiple and thorough assessments to determine a candidate's motivation, social disposition, and emotional intelligence before preparation courses may better identify candidates and align potential job and organization fit within the discipline. Practical Application: Applications of the findings would include a standardized selection process with improved interviews and pre-course auditing, and candidate expectation management before the selection to attend preparation or certification courses. The efforts potentially decrease long-term costs and deficiencies when candidates have an inconsistent job or organizational fit, departing from organizations after short periods or by not providing consistent quality instruction to students. The study recommendations, when implemented, can improve most educational disciplines where instructors are selected for technical instructional positions where students risk injury or harm.

A thousand years of crash experience in three hours: An online hazard perception training course for drivers

A key goal of driver training is to teach drivers to avoid crashes. However, in traditional driver training, drivers are unlikely to see even a single example of the class of event that we want them to learn to avoid. We developed a six-session automated online hazard perception training course for drivers, which incorporates a range of evidence-based strategies and employs extensive video footage of real crashes. We evaluated this course in a randomized control trial by examining its effects on previously-validated computer-based measures of hazard perception, hazard prediction, speed choice, following distance, and gap acceptance propensity, as well as self-rated measures of driver skill, safety, and real world transfer. We found that the course resulted in significant improvements in hazard perception response time and hazard prediction scores, and significantly longer vehicle following distances. Additionally, all participants in the trained group reported that their real world driving behaviour had improved. No significant training effects were found for the other measures. The results suggest that the course can improve key behaviours associated with crash risk.

The development and validation of video-based measures of drivers' following distance and gap acceptance behaviours

The distance at which drivers follow other vehicles has been found to be linked to crash risk. Tailgating (i.e. driving at an unsafe following distance) is both endemic and a leading cause of rear-end crashes. Similarly, drivers' decisions about when to merge with a stream of traffic are likely to influence crash risk. Consistent with this, it has been shown that crashes are more common at intersections where drivers more frequently have to slow for vehicles pulling out into insufficient gaps. Therefore, the development of reliable and valid measures of both of these driving behaviours would facilitate further crash prevention research. Given the problems associated with assessing these behaviours during real driving, we developed new video-based measures. In our new following distance measure, participants view videos shot from the perspective of a driver who is following another vehicle at a range of distances across a variety of traffic environments. On each trial, participants report their own minimum comfortable following distance relative to the following distance depicted in the video. In our new test of gap acceptance behaviour, participants view a series of video clips and indicate when they would pull out into the approaching stream of traffic shown in each clip. The two new measures each yielded reliable data, and we found that young drivers made riskier choices than older drivers for both following distance and gap acceptance. These age-related differences are consistent with those found in observational studies of real driving, supporting the proposal that the new tests could potentially be used as proxies for these crash-related driving behaviours in both lab-based research and large-scale online studies.

Road user hazard perception tests: A systematic review of current methodologies

BACKGROUND

Poor hazard perception, or the ability to anticipate potentially dangerous road and traffic situations, has been linked to an increased crash risk. Novice and younger road users are typically poorer at hazard perception than experienced and older road users. Road traffic authorities have recognised the importance of hazard perception skills, with the inclusion of a hazard perception test in most Graduated Driver Licensing (GDL) systems.

OBJECTIVES

This review synthesises studies of hazard perception tests in order to determine best practice methodologies that discriminate between novice/younger and experienced/older road users.

DATA SOURCES

Published studies available on PsychInfo, Scopus and Medline as at April 2018 were included in the review. Studies included a hazard perception test methodology and compared non-clinical populations of road users (car drivers, motorcyclists, bicyclists and pedestrians), based on age and experience, or compared methodologies.

RESULTS

49 studies met the inclusion criteria. There was a high degree of heterogeneity in the studies. However all methodologies - video, static image, simulator and real-world test-drive were able to discriminate road user groups categorised by age and/or experience, on at least one measure of hazard perception.

CONCLUSIONS

Whilst there was a high level of heterogeneity of studies, video methodology utilising temporal responses (e.g. press a button when detecting the potential hazard) are a consistent measure of hazard perception across road user groups, whereas spatial measures (e.g. locate potential hazard in the scenario) were inconsistent. Staged footage was found to discriminate as well as unstaged footage, with static images also adding valuable information on hazard perception. There were considerable inconsistencies in the categorising of participants based on age and experience, limited application of theoretical frameworks, and a considerable lack of detail regarding post hoc amendments of hazardous scenarios. This research can guide further developments in hazard perception testing that may improve driver licensing and outcomes for road users.

How Can On-Road Hazard Perception and Anticipation Be Improved? Evidence From the Body

The present research is aimed at investigating processes associated with learning how to drive safely. We were particularly interested in implicit mechanisms related to the automatic processing system involved in decision making in risky situations (Slovic et al., 2007). The operation of this system is directly linked to experiential and emotional reactions and can be monitored by measuring psychophysiological variables, such as skin conductance responses (SCRs). We focused specifically on the generalization of previously acquired skills to new and never before encountered road scenarios. To that end, we compared the SCRs of two groups of participants engaged, respectively, in two distinctive modes of moped-riding training. The active group proceeded actively, via moped, through several simulated courses, whereas the passive group watched video of the courses performed by the former group and identified hazards. Results indicate that the active group not only demonstrated improved performance in the second session, which involved the same simulated courses, but also showed generalization to new scenes in the third session. Moreover, SCRs to risky scenes, although present in both groups, were detectable in a higher proportion in the active group, paralleling the degree of risk confronted as the training progressed. Finally, the anticipatory ability demonstrated previously (and replicated in the present study), which was evident in the repeated performance of a given scenario, did not seem to generalize to the new scenarios confronted in the last session.

A First Step toward the Understanding of Implicit Learning of Hazard Anticipation in Inexperienced Road Users Through a Moped-Riding Simulator

Hazard perception is considered one of the most important abilities in road safety. Several efforts have been devoted to investigating how it improves with experience and can be trained. Recently, research has focused on the implicit aspects of hazard detection, reaction, and anticipation. In the present study, we attempted to understand how the ability to anticipate hazards develops during training with a moped-riding simulator: the Honda Riding Trainer (HRT). Several studies have already validated the HRT as a tool to enhance adolescents’ hazard perception and riding abilities. In the present study, as an index of hazard anticipation, we used skin conductance response (SCR), which has been demonstrated to be linked to affective/implicit appraisal of risk. We administered to a group of inexperienced road users five road courses two times a week apart. In each course, participants had to deal with eight hazard scenes (except one course that included only seven hazard scenes). Participants had to ride along the HRT courses, facing the potentially hazardous situations, following traffic rules, and trying to avoid accidents. During the task, we measured SCR and monitored driving performance. The main results show that learning to ride the simulator leads to both a reduction in the number of accidents and anticipation of the somatic response related to hazard detection, as proven by the reduction of SCR onset recorded in the second session. The finding that the SCR signaling the impending hazard appears earlier when the already encountered hazard situations are faced anew suggests that training with the simulator acts on the somatic activation associated with the experience of risky situations, improving its effectiveness in detecting hazards in advance so as to avoid accidents. This represents the starting point for future investigations into the process of generalization of learning acquired in new virtual situations and in real-road situations.

The effect of performance feedback on drivers' hazard perception ability and self-ratings

Drivers' hazard perception ability has been found to predict crash risk, and novice drivers appear to be particularly poor at this skill. This competency appears to develop only slowly with experience, and this could partially be a result of poor quality performance feedback. We report an experiment in which we provided high-quality artificial feedback on individual drivers' performance in a validated video-based hazard perception test via either: (1) a graph-based comparison of hazard perception response times between the test-taker, the average driver, and an expert driver; (2) a video-based comparison between the same groups; or (3) both. All three types of feedback resulted in both an improvement in hazard perception performance and a reduction in self-rated hazard perception skill, compared with a no-feedback control group. Video-based and graph-based feedback combined resulted in a greater improvement in hazard perception performance than either of the individual components, which did not differ from one another. All three types of feedback eliminated participants' self-enhancement bias for hazard perception skill. Participants judged both interventions involving video feedback to be significantly more likely to improve their real-world driving than the no feedback control group. While all three forms of feedback had some value, the combined video and graph feedback intervention appeared to be the most effective across all outcome measures.

Hazard Perception in Driving

Hazard perception in driving refers to a driver’s ability to anticipate potentially dangerous situations on the road ahead and has been the subject of research for over 50 years. It is typically measured using computer-based hazard-perception tests and has been associated with both retrospective and prospective crash risk, as well as key crash-risk factors such as distraction, fatigue, alcohol consumption, speed choice, and age-related declines. It can also differentiate high- and lower-risk driver groups. The problem is that it is also a skill that appears to take decades of driving experience to acquire. This raises the question of whether it is possible and practical to accelerate this learning process via assessment and training in order to improve traffic safety. We have evidence that, in contrast to most driver education and assessment interventions, hazard-perception testing and training appear to have the capability to reduce crash risk. For example, the inclusion of a hazard-perception test in the UK driver licensing process has been estimated to reduce drivers’ non-low-speed public-road crash rates by 11.3% in the year following their test.