Modeling pedestrian movement in shopping street segments

The main purpose of this paper is to test a model that aims at modeling and simulating micro pedestrian behavior in shopping street segments, including entering shops. The model assumes a detailed network of links to represent the structure of street segments and entrances to the shops. The basic principle underlying the model is that a pedestrian chooses a destination when entering a shopping street segment. Destinations may be shops located along the segment or the exit points at the opposite side of the segment. The choice of a destination is modeled by means of a discrete choice model, including variables such as type-specific supply of shops, distances, and tendency to visit a shop. After choosing a destination, the route to that destination is modeled by a discrete choice model as well. However, route choice is modeled at the link level, not at the level of complete routes. A sequence of chosen links constitutes the route to the destination. Relevant variables included in this model are walking distance, and angles between links and the direction of the destination. If the destination is a shop, the destination choice model is applied again to select the next destination. This process is repeated until one of the exit points of the segment is chosen or the pedestrian stays in a shop. The study area is the main shopping street of Antwerp (Belgium). During a one-week workshop in July 2004, students observed pedestrian movement in this shopping street. An inventory of physical characteristics of the shopping street was made and pedestrians were unobtrusively tracked through two separate segments of the shopping street (approximately 100 m of length each). In total, 335 pedestrians were tracked. The choice models were estimated using the tracked routes. Next, the models were used to simulate pedestrians' behavior by consecutively selecting destinations and links to these destinations. The models perform well. Observed and simulated routes were used to determine and compare observed and simulated link loadings. It can be concluded that by using observational data only, relevant models of pedestrian behavior can be developed.