Simulating pedestrian flow through narrow exits

Abstract

Abstract Introduction: The flow of pedestrians through narrow doorways is one of the most common features of crowd motions and evacuations. It is particularly an important aspect of pedestrian simulations models since their accuracy depends highly on their ability to produce realistic exit flow rates. The problem has been extensively studied in the literature, but many aspects of it have remained controversial with mixed (and often contradictory) evidence emerging from different studies and different methods. Methods: We discuss the significance of parameter calibration for accurate simulation of pedestrian flow through narrow exits using social force model. Based on sensitivity analyses, we show how simulated exit throughput rate can vastly differ by changing the value of certain parameters. We identify the two parameters that are most critical, and then calibrate them based on a set of experimental observations (at macro level). Using these calibrated parameters, we then re-examine three fundamental questions related to pedestrian flow at bottlenecks, (1) the relation between desired velocity and simulated egress time; (2) the effect of barricade at exits; and (3) the effect of exit in the corner versus the middle. Results: Our numerical analyses showed that, with the calibrated parameters, increasing the desired velocity in the social-force model results in monotonically shorter egress times (at a marginal rate that rapidly diminishes as the desired velocity increases). We showed that placing a panel-like barricade at exit can facilitate the outflow and reduces the egress time, but its effect depends on the widths of exit, as well as the size of the barricade and its distance to exit. We show that the positioning the exit in the corner is also effective in terms of reducing egress time, but only for very narrow exits. The benefit diminishes quickly as the exit becomes wider. Applications: These outcomes demonstrated the significance of parameter calibration for accurate simulation of crowd flows. The findings may also help to identify simple modifications that can facilitate crowd flows at narrow bottlenecks.