Simulating Car-pedestrian Interactions During Mass Events with DTA Models: the Case of Vancouver Winter Olympic Games



Simulating Car-pedestrian Interactions During Mass Events with DTA Models: the Case of Vancouver Winter Olympic Games

Authors

L Meschini, G Gentile, DITS - Sapienza Università di Roma, IT; K Lew, PTV, US

Description

This paper presents the within-day Dynamic Traffic Assignment model implemented to simulate ordinary, evacuation and emergency scenarios for downtown Vancouver during the forthcoming Winter Olympic Games.

Abstract

This paper presents the within-day Dynamic Traffic Assignment model implemented to simulate ordinary, evacuation and emergency scenarios for downtown Vancouver during the forthcoming Winter Olympic Games. Within this context several models are proposed and analyzed to reproduce different kinds of pedestrians and vehicles-pedestrians interactions. These congestion phenomena can occur in presence of the unusual demand produced by important public events, such as sport games, music concerts and political rallies, when significant levels of pedestrian and vehicle flows are concentrated in space and time, i.e. converge to or diverge from one point/area in a relatively short interval.
Among the existing approaches that satisfy the above identified simulation needs, we have considered the DUE model available in VISUM (Gentile et al., 2006). In order to apply this method, an existing macroscopic flow model (Gentile et al., 2005) has been suitably extended to represent the interactions between vehicles and pedestrians in presence of explicit capacity constraints on roads and sidewalks. Developing the model proposed by Daamen and Hoogendoorn (2003), who investigate the experimental relation between the longitudinal space used by pedestrians and their speed, we have specifically addressed the problem of correctly allocating to each direction the capacity of a sidewalk used by opposite pedestrian flows, showing how this is an endogenous variable that results from the assignment itself.
Five different types of vehicle-pedestrian interaction are identified, and a model is devised for each one of them.
? normal interaction ? occurs when pedestrians use only sidewalks and cross roads at specific points. In this scenario, no longitudinal interaction takes place, while transversal interaction can be neglected or taken into account through a priori turn delays for cars. This case can be addressed performing two separate DTA (pedestrian and car) in no particular order.
? controlled interaction ? occurs during special events, when some road lanes can be assigned to pedestrians. While we can still reasonably assume that no longitudinal interaction occurs, transversal interaction may not be negligible, due to the high number of pedestrians involved. However, it would be plausible in this case that pedestrians have priority over car when crossing roads, which means that turn delays and turn capacities for cars depend on pedestrian flows at intersections. This case can be modelled performing a pedestrian DTA with increased capacities, then based on those results performing a car DTA with turn delays and turn capacities accordingly with pedestrian flows at intersection and crossing points, through suitable delay and capacity reduction functions.
? random interaction ? occurs during special event and/or evacuation scenario, when pedestrians may randomly and discontinuously occupy part of the car lane, causing a sort of ?longitudinal friction? with car flows. We think here to a situation with very crowded sidewalks, where faster pedestrians ?hop-off? and ?hop-on? the sidewalk in order to pass slower people. This phenomenon leads to a reduction of the road capacity and free-flow speed; drivers have indeed to reduce their speed for safety reasons, because pedestrians may randomly interact with their trajectory. This case can be modelled performing a pedestrian DTA with increased capacities, then based on those results performing a car DTA with reduced car link free flow speeds, capacities and number of lanes accordingly with pedestrian flows on the same link.
? chaotic concordant interaction ? occurs in presence of evacuation scenarios, when it is not possible to separate pedestrian and car flows, which are completely mixed but travel in the same direction. This assumption may be reasonable if cars and pedestrians are trying to reach the same collecting point. In this case, the sidewalk capacity is assigned to pedestrians, while the entire road capacity is assigned to both car and pedestrians, where cars are thus forced to travel at pedestrian speed. This scenario can be simulated with a unique multi-modal DTA, where cars are considered as ?pedestrian equivalents?.
? chaotic conflicting interaction ? is the worst case, occurring in presence of evacuation scenarios, when pedestrian are mixed with cars and do not respect their direction. In this case, cars are basically stuck on the road, while pedestrians flow around them. We assign to pedestrians sidewalks, plus a small part of the road capacity. Once roads are cleared from pedestrians cars can move and evacuate.
Finally, the proposed models are investigated by means of a thorough numerical analysis, whose results are presented and discussed, both on simple networks and for the real case of downtown Vancouver.

Keywords: car-pedestrian congestion, dynamic traffic assignment, important event simulation, Vancouver Olympic Games.

Publisher

Association for European Transport