The authors implemented a novel visual tool to analyse driver behaviour in simulated car crashes. The crashes occurred in high-fidelity collision avoidance scenarios staged on the Iowa Driving Simulator. Driver performance data were digitised at 30 Hz and crash events were marked, allowing to "rewind" the data stream to analyse driver performance in the final moments. The authors were able to produce diagrams that clearly disclose the "anatomy" of a crash. This was made possible by plotting vehicle speed, lane position, brake, accelerator and steering wheel position, and driver gaze position in the 5 seconds prior to collision, together with a representation of road geometry and trajectory of neighbouring vehicles. A better understanding of the human errors that cause crashes can ultimately aid in the development of more rational strategies for crash prevention. (A)
Samenvatting