Safety implications of driver misperception in road accidents involving child pedestrians
Accidents to child pedestrians are usually blamed on their inexperience and carelessness. This thesis proposes instead that the main problem is a form of driver error, overestimation of time-to-collision. When drivers have to make critical decisions about braking, time-to-collision may be much longer than can be reliably judged from optic flow. They therefore have to obtain time-to-collision indirectly from distance and speed. The main cue for distance is familiar size, so if a driver tends to misperceive a child as a larger person at a greater distance, he would overestimate time-to-collision. Risk of accident would therefore increase. That hypothesis is supported by a new collision simulation, which replicates a braking manoeuvre. An experiment based on the simulation supports the hypothesis, and allows influences on accidents to be predicted. These are tested from national accident statistics, which indicate that over half of all child pedestrian casualties would not occur but for this driver error. Several remedial measures are proposed. The experiment not only supports that hypothesis. It also suggests that the cognitive process by which time-to-collision is obtained from optic flow differs from that generally assumed - processing angular distance and angular velocity by the algorithm θ/θ. An alternative algorithm 2theta/θis proposed, where θ is angular acceleration, which provides the same value of time-to-collision without the need for angular distance. The new algorithm is supported by further experiment, and provides an understanding of situations which were difficult to reconcile with the θ/θ hypothesis. Our ability to respond to moving objects which are very small, of variable shape, visible only briefly, or changing speed, becomes more understandable. Though primarily related to pedestrian safety, therefore, the work reported in this thesis could have far wider application.