Automated human choice extraction for evacuation route prediction

This paper introduces a new Automated Cue Choice Extraction Method (ACCEM), which is designed to automatically deduce subject's choices of visually perceivable architectural cues from their trajectories observed in a VR-based evacuation experiment. ACCEM is developed to generate input for evacuation simulation modeling. Complementary to the existing video-based automated analysis on crowd evacuation, ACCEM utilizes a Virtual Reality (VR) based evacuation experiment to automatically analyze the individual visual perception as input. First, the computational model of ACCEM is introduced. Next, the vision-based recognition algorithm and the reverse reasoning in ACCEM are explained in detail. Last, a demonstration of ACCEM successfully deducing the subject's choices along an observed trajectory is illustrated. After discussing the application issues of ACCEM, it is concluded that ACCEM fits for its purpose and that it provides a new opportunity to VR-based evacuation research.