Can a driving simulator assess the effectiveness of Hazard Perception training in young novice drivers?

Hazard awareness skills of young-novice drivers are receptive to training. Yet, there is no consensus on an optimal training program or acceptable measures to assess intervention effectiveness. The current study aimed to develop a driving Simulator based Hazard Perception Test (S-HPT) for assessing changes in awareness of hazards among trained young-novice driver, relative to a control group, and a group of experienced drivers (gold standard). Two urban environments were chosen for testing; residential district with narrow roads, and business district with wider roads. Fourteen scenarios consisting of potential situations known to differentiate between untrained young-novice and experienced drivers (e.g., limited field of view) with and without materialized situations were created. Data from 39 young-novice and 6 experienced drivers was analyzed. Using an allocation algorithm each young participant was assigned to one of three training interventions (AAHPT-Active, Hybrid or RAPT) or control. Driving speed was chosen as the measure of analysis since it indicates whether drivers have spotted a hazard, and what they chose to do to avoid it. It was sampled every 3 meters, approximately 400 times per scenario per driver group. A respective curve was fitted for each group. Differences in speed among driver groups were examined and comparisons between potentially hazardous scenarios and materialized ones were made. Results revealed that scenarios that required braking but not to a complete stop were the most diagnostic. The use of two environments demonstrated differences in group behaviours that would have not been seen if utilizing only one environment. Use of two comparisons, to control and to experienced drivers was reassuring as the results complemented. Where the resemblance of trainees was higher to control, they tended to resemble the experienced group less. The Active trainees bore more resemblance to the control in the residential scenarios while the RAPT and Hybrid bore more resemblance to the control in the business district scenarios. To conclude, generation of a variety of scenes in multiple traffic environments with- and without materialized hazards is a promising way for hazard awareness simulator test development. Group-related metrics and analyses are essential to investigate which conditions discriminate among driver groups.