Modeling of pedestrian behavioral dynamics in complex urban settings and built environment

Abstract

The movement of people makes a place vibrant. Thus, pedestrian movement patterns through space must be considerably understood by transportation scientists, urban designers, and retailers. However, given heterogeneous and complex properties, how pedestrian behaviors can be realistically reproduced remains as an open issue. Computer simulation models provide a potential means by which the interactions between pedestrians and different spatial environments can be revealed at the microscopic level. Although some attempts on modeling and simulating pedestrian movement have been made for decades, a lack of appropriate tools to realistically represent pedestrian behaviors is experienced. A hybrid agent-based pedestrian behavior simulation model—CityFlow-U was developed in this study to reproduce varied realistic pedestrian movement behaviors in complex urban settings and built environment. The section-and-portal graph approach was proposed to facilitate pedestrian navigation in the environment, in which the sections were divided into four subsets according to neighborhood relationship and linked portal characteristics. Similar to most simulation models, the model in this study was implemented at three different levels, namely, the strategic level, the tactical level and the operational level. Four major modules that focused on pedestrian behavior at different circumstances and levels were designed and developed. Elements at the strategic level dealt with trip nature determining, activity scheduling and time budgeting, that is, initiation and termination of a pedestrian trip and the decisions on making unplanned activities. At the tactical level, two different routing strategies were considered. One was The Goal-Oriented Static Navigation Module that utilized under situations wherein pedestrians were familiar with the environment and had specific destinations, while the other was The VGA-Based Dynamic Navigation Module that dynamically investigated situations wherein pedestrians had no specific destinations. Results of the Visibility Graph Analysis (VGA) of the spatial environment were innovatively used as an important reference for pedestrian’s dynamic routing decisions as it has been reported that pedestrians’ general movements were directed along their lines of sight. Along with other significant factors, The VGA-Based Dynamic Navigation Module reproduced realistic trails and microscopically generated natural movements. By comparing simulation results of pedestrian movements with empirical spot count and trail data in Tate Gallery obtained from previous literatures, the said module proved to be reliable and flexible. The implementation of pedestrian locomotion behaviors at the operational level were achieved by The Goal-Directed Locomotion Module and The Exploratory Locomotion Module. In the former module, aside from common behaviors that include steering and collision avoidance, emergent behaviors such as attracted movement and side-by-side walking group were considered, illustrated explicitly and reproduced with considerable credibility. Nonetheless, a series of testing under a notional scenario proved that the formation and maintenance of these behaviors posed negative influence on pedestrian dynamics. Innovatively, The Exploratory Locomotion Module was developed to reproduce pedestrian exploratory behavior, by which a rarely explored activity choice behavior, impulse stop, was successfully reproduced using simple heuristics. Moreover, observational studies on pedestrian exploratory behavior in a shopping mall were carried out, and the credibility of the module was demonstrated by the good uniformity of simulated impulse stop behavior and data in the empirical video.

Date of Award	16 Feb 2015
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Siu Ming LO (Supervisor)