Julia Ferri - Fire and Civil Engineering, York University
Crowd simulation software is an emerging technology used to validate and verify stadium design in terms of total egress of large crowds. To model occupant movement, this technology requires an array of input variables assigned by the user. Currently these inputs lack diversity in movement representation. Limited studies have worked towards developing walking speeds for different demographics; however, none are project-specific to stadium design nor provide usable statistics regarding physical impairments and obesity, for example. Models are therefore challenged in their ability to diversify crowd demographics and represent realistic evacuations. With realism not assured, there are a range of uncertainties that can be introduced by assumptions from the user. To overcome these limitations, this research seeks to perform an analysis of the demographics seen at a stadium, establish a set of walking speeds, and use this to authenticate available egress simulation models.
High-resolution cameras were used to record pedestrian circulation and egress at a Canadian stadium during a sports tournament in 2019. The footage was later analysed to qualitatively assess behavioral patterns and develop walking speeds for the following profiles: children, young adults, adults, seniors, families, overweight adults, overweight seniors, cane, crutches, wheelchair (manual and electric), mobility scooter, walking stick, carrying oversize luggage, and requiring assistance. This data was then used in crowd simulation software (based on a commonly utilised modified social forces framework) to compare with earlier modelling methods against a recorded evacuation of the stadium.
Individualizing profiles in total egress modelling provides a step towards reducing uncertainties of human behavior and producing more reliable frameworks for crowd movement predictions. The importance of diversifying input speed parameters is revealed in comparison to previously relied upon methods that limit crowd behavior to a single range of movement, and additionally, advocates for project-specific data acquisition. Limitations of the final model are discussed, and suggestions are made for continued studies on movement behavior and improvements to current software.