PubMed2003
Road traffic accidents are the single greatest cause of fatality in the workplace and the primary cause of all accidental death in the US to the age of 78. However, behavioural analysis of response in the final seconds and milliseconds before collision has been a most difficult proposition since the quantitative recording of such events has largely been beyond cost feasibility for road transportation. Here, a new and innovative research strategy is reported that permits just such a form of investigation to be conducted in a safe and effective manner. Specifically, a linked simulation environment has been constructed in which drivers are physically located in two adjacent, full-vehicle simulators acting within a shared single virtual driving world. As reported here for the first time, this innovative technology creates situations that provide avoidance responses paralleling those observed in real-world conditions. Within this shared virtual world 46 participants (25 female, 21 male) were tested who met in two ambiguous traffic situations: an intersection and a hill scenario. At the intersection the two drivers approached each other at an angle of 135 degrees and buildings placed at the intersection blocked the view of both drivers from early detection of the opposing vehicle. The second condition represented a 'wrong' way conflict. Each driver proceeded along a three-lane highway from opposite directions. A hill impeded the oncoming view of each driver who only saw the conflicting vehicle briefly as it crested the brow of the hill. Driver avoidance responses of steering wheel, brake, and accelerator activation were recorded to the nearest millisecond. Qualitative results were obtained through a post-experience questionnaire in which participants were asked about their driving habits, simulator experience and their particular response to the experimental events which they had encountered. The results indicated that: (1) situations have been created which provided avoidance responses as they have been recorded in real-world circumstances, (2) the recorded avoidance responses depended directly upon viewing times, and (3) the very short viewing times in this experiment resulted in a single avoidance action, largely represented by a random choice of swerve to either right or left. The present results lead us to posit that in order to be able to design accident avoidance mechanism that respond appropriately in the diverse situations encountered, there is a need to pay particular attention to mutual viewing times for drivers. The general implications for a behavioural science of collision-avoidance are evaluated in light of the present findings.