Line-of-sight Pedestrian Microsimulation in Urban Systems

In order to move away from more formal route-choice models and to avoid calibration issues, visually guided agent-based models have recently been introduced to microscopic urban models of pedestrian behavior. Although the new models produce good correlation between the agents and observed pedestrian movement, there are two problems with them. Firstly, the implementations have lacked an origin-destination matrix, and so the models work best in situations where undirected exploratory movement is likely to occur. Secondly, the models rely on precalculated visibility graphs to store visual data from locations within the system; these graphs use a large amount of computer memory. Here we show that a line of sight model using reduced visual information compares well with the destination selection method, and further, that the addition of a simple origin-destination matrix can lead to better agent correlation with pedestrian movement.