Abstract:
The increasing frequency of mass-gathering events in recent years poses the essential need for realistic simulation and prediction of crowd movements, especially in crowded situations, before implementing further applications like risk assessments or safety control. However, different crowd phenomena have been observed at certain levels of density while each modeling approach typically captures one or a few specific phenomena. This talk discusses the use of density-related factors in hybrid approaches for two primary focuses: coupling pedestrian modeling approaches to cover a wider range of crowd phenomena and enhancing the realism of pedestrian trajectory predictions using deep learning in crowded scenarios.