The experiment brought together 120+ naïve drivers, the largest number of participants ever gathered in a dynamic driving simulator for a test on the characterization of drivers’ visual distraction patterns.
The Driver Visual Distraction Characterization (DVDC) experiment, conducted in collaboration with Toyota Motor Europe, investigated and assessed driver visual attention associated with road traffic accidents and near-misses. The characterization of visual distraction patterns in drivers was based on a study at Applus+ IDIADA’s driving simulator laboratory with 130 volunteer drivers. Results of this large-scale study were presented at the 7th International Conference on Driver Distraction and Inattention, which took place in Lyon, France, in October 2021.
Short visual distractions involve secondary tasks, taking drivers’ attention away from the primary driving task. These reduce the driver’s capability to effectively respond to sudden changes in the road environment, and closely relate to crucial errors leading to a higher accident risk rate. Given their prevalence, risks associated with visual distraction are a key focus in road safety.
Much investigation has been conducted into the classification of tasks contributing to visual distraction in drivers and they are widely reported. Nevertheless, objective parameters that may be used in the classification of visual distractions and their associated risks are less well defined. Therefore, it is important to characterize visual distractions by objective criteria and evaluate their impact on the risk of accidents under relevant road scenarios.
The study used a multi-stage approach methodology to investigate drivers’ distractions and associated risks leading to road accidents. The first stage evaluated critical scenarios associated with visual distractions by means of literature-based accidentology studies, complemented with accident database statistics. Rear-end, crossing and single vehicle events were identified as the most relevant scenarios, having a high frequency of distraction-related accidents and a higher degree of severity.
The second stage mostly focused on studying drivers’ visual distraction behaviour by general task and objective components. Visual distraction was defined with the aim of forming a taxonomy of visual distraction behaviours in drivers. It was conducted through an examination of existing literature around visual distraction characteristics and associated secondary tasks, and a naturalistic field study involving seven volunteer drivers over 1.164 km. The principle collected metrics came from gaze detection of target zones away from forward view, alongside a video feed of the driver, and collection of vehicle data metrics.
In parallel, participants were asked to provide subjective feedback regarding prevalent secondary tasks, involving visual interaction with targets not associated with the driving task. Participants were asked to identify and rank each of these. Outcomes were used in defining representative driver distraction inducements, which were defined by the combination of the visual task interaction (timesharing or fixation) and its location in the vehicle.
Both stages subsequently culminated in a driver simulator study which involved 126 naïve driver participants. Twelve conditions were tested, with four visual task inducements (see picture 3) representative of identified distraction characteristics, and three identified critical scenarios (see picture 1). Drivers were instructed to engage in one inducement task at multiple defined points whilst driving in the simulated environment. During the distraction inducement, each of the critical scenarios was presented to them. Data was collected regarding scenario outcome, driver’s state of visual attention, driver’s reaction, and driver’s perception of the scenario severity and task difficulty.
Rear-end event and crossing scenario outcomes were assessed according to the overall outcomes regarding whether there was a collision or not. Supporting assessment was conducted with regard to collision severity by a collision-speed metric, with driver reaction times also being used.
Rear-end event task B (fixation visual task with a low console screen position – see picture 3) showed overall highest instance of collision outcome (see picture 4). Higher severity of this task was supported by collision speed and reaction time impacts. Another conclusion was that there were far fewer instances of collision in timesharing tasks combined with a position further away from a forward view associated with higher risk outcomes.
However, in crossing event task C (timesharing visual task with a low console screen position – see picture 3) showed the highest instance of collision outcomes and timesharing task for same position, C and B, showing higher risk than fixation (see picture 5). All of this according to collision speed results, and for driver reaction time.
There were no collisions to assess among all the voluntary drivers encountering a single vehicle event, so the evaluation of outcomes assessed how able the drivers were to keep the vehicle in lane and under control, supported by statistics for deviation in vehicle heading relative to road direction. Task B had the most severe trajectory deviation outcomes for single vehicles, both in task position and task type (see picture 6).
Finally, general outcomes showed association of a higher risk when the secondary fixation task visual target was positioned further away from the forward view, as evidenced in a direct comparison between tasks A and B, which was consistently more severe for the lower positioned target B.
When comparing between the locations of the timesharing tasks, C and D, the impact of position was less clear. No evidence was found when trying to identify an underlying cause. However, when considering the position of elements within the critical event, it may be possible that this could impact the drivers’ capability to perceive the critical threat within a scenario. This may well explain the variability in the obtained outcomes.
Furthermore, evidence found that generally supported timesharing tasks were associated with lower risk outcomes when compared to fixation tasks’ behaviour. Only in the case of crossing event this was not consistent, since it demonstrated a higher risk associated with timesharing tasks. It was not possible to identify any causational factors for this effect.
The outcomes have provided valuable data and it has been made evident that this could be a relevant area for further future investigation regarding the assessment of the effect of distraction target position relative to road traffic scenarios, and overall accident risk.
The automotive industry is at a crucial point in the development of new and more complex driver assistance systems, especially as Euro NCAP is actively working on new regulations that will have a considerable impact on the efficiency and robustness requirements of new ADAS systems.
Applus+ IDIADA is working on the characterization of adverse states and on the development of test protocols to assess them, having already developed some very efficient ones. All of this, thanks to a team of Human Factors specialists with extensive experience, as well as a state-of-the-art Dynamic Driving Simulator and eye tracking systems which can help to offer DVDC testing and development services adapted to the needs of our clients.
Davide Salanitri, Project Engineer at Applus+ IDIADA and PhD in Human Factors, said “this project was a big step forward in the characterization of visual driving behaviour, the most common types of distractions among drivers. This line of studies that we conducted will allow us to help our clients develop a more efficient system and, overall, a safer vehicle”.
Cristina Periago, Psychologist, and member of the Human Factors team at Applus+ IDIADA, added “it was a real challenge testing with more than 120 naïve volunteering drivers and observing their visual behaviour during simulated driving, such a big sample has provided a large amount of interesting data that will undoubtedly be useful for future monitoring systems development”.
Jordi Bargalló, ADAS Manager at Applus+ IDIADA, further added: “We want to thank Toyota Motor Europe for entrusting us with such an insightful experiment that will surely add to valuable data of drivers' behavioural patterns in moments before and during situations of visual distractions that can lead to traffic accidents”.
On behalf of Toyota Motor Europe, Maminirina Ranovona, Safety Research and Technical Affairs Manager, stated “we are satisfied with the data obtained, which will undoubtedly be of great value in our work to further improve ADAS systems. Insight gained from this data could help develop ADAS systems that are capable of detecting anomalies in driver behaviour and, thereby, preventing dangerous driving situations”.