Liliana PEREZ, Université de Montreal, Canada
Jeffrey KATAN, SYSTRA Canada, Canada
Joaquin ORTIZ, SYSTRA Canada, Canada
Crowd movement is a complex scenario with serious implications in urban planning, architecture, and security. Moreover, individual movement and crowd behaviour is an important subject in many research domains, including geography, traffic engineering, psychology, robotics, and video games. Despite decades of research into crowd simulation modelling, certain challenges pertaining to some modelling approaches persist, and may even be underrepresented. Agent-based models for crowd simulation often use navigation fields for tactical pathfinding because it greatly simplifies path calculation and scales well with large numbers of agents. Navigation fields are often combined with separate operational movement models such as the Social Force Model to determine agent behaviour in response to obstacles and other nearby agents. However, our research has revealed that the independent validity of these operational and tactical models may not necessarily scale up when combined in scenarios with widely varying geometries. Base cases are a common benchmark of complex, emergent patterns for crowd simulation model validity, but such validity may not hold up against the complexities of scenarios with multiple representations of these base cases. Furthermore, navigation fields can be scale-dependent, with varying sensitivity over distance.
Mots clés : Navigation field|agent-based modelling|crowd movement|complex systems|scale issues