In the early days of pedestrian micro-simulation, the flow of the individuals did not look natural to an observer;
the movements were very jumpy and did not appear to replicate the expected flows of pedestrians that are observed
in crowds. To overcome this jumpiness, modelers introduced coherent and smoothing functions. The two most
famous of these are Reynold’s Boids model (Reynolds (1987)) and the Social Forces of Helbing and Molnar (1995).
Both of these approaches average out the potential influences on a simulated individual’s heading and produce an
aesthetically pleasing visualization of the flow of pedestrian movement. One reason for the improved results is that
all of the simulated pedestrians tend to have the same directional goals; that is, they are homogeneous. What if the
simulated pedestrians were heterogeneous and had different heading objectives like maintaining cohesion within
their own individual groups as well?
In this study, the authors use Repast Simphony to build an agent-based model (ABM) that incorporates group
cohesion forces into this type of pedestrian egress scenario. The pedestrian agents have two goals: (1) exit the venue
and (2) maintain a level of cohesion within their group. The study compares two approaches to updating agent
headings while incorporating these two goals. The first approach uses the weighted averaging approach of social
forces (Helbing and Molnar (1995)). The second approach considers a discrete stochastic selection approach
between the two goals, which is updated at each time-step. The results from this approach were quite surprising. The
paper also considers the impact of the level of group cohesion.
Several other researchers have looked at the effects of group dynamics on simulated pedestrian egress (Vizzari et
al. (2013), Wijermans et al. (2013), Pluchino et al. (2014)). However, other studies used a weighted average heading
approach to dealing with multiple individual pedestrian objectives. We believe that this is the first paper that takes a
step-back from the currently accepted approach of weighted averaging headings to consider ways to overcome its
limitations. The approach to group dynamics, outlined in this paper, is a simple one for this simple purpose; for a
more detailed approach to pedestrian group dynamics, please see Elzie et al. (2014).
The next section discusses the scenario and the model design. This is followed by a discussion on the simulation
results. Finally, conclusions are given.