Simulating dynamics of adaptive exit-choice changing in crowd evacuations: Model implementation and behavioural interpretations

摘要

A crucial aspect of simulating crowds' evacuation processes is that humans can dynamically revisit and change their decisions. While a relatively great deal of attention has been paid by recent studies to modelling directional decision making, the 'exit decision changing (or decision adaptation)' phenomenon has been largely overlooked. Here, we quantitatively investigate (I) how important is to include a decision changing module in evacuation simulation models, and (II) whether decision changing is beneficial to evacuation processes. We propose and implement a parsimonious discrete-choice model of decision changing. The model embodies the most influential factors that make an evacuee revise and adapt their choice of exit. This includes the effects of 'relative queue-size imbalance at exits', 'visibility of exits', 'social influence' and 'inertia (for maintaining initial decisions)'. Results showed that, the inclusion of the decision changing module made a very substantial difference in enhancing the accuracy of the simulation outputs. Simulating exit choices as one-off decisions strictly limited the degree of match that could be achieved between the simulated and experimental outputs (in terms of replicating the observed exit shares and evacuation times) (question I). Further analyses also revealed that an intermediate degree of decision changing is a strategy that most benefits the system. By contrast, the extreme decision-changing strategies (i.e. "no change" and "too many changes") were found to be suboptimal. Also, while we have observed, in our other studies, that imitative (or the so-called herd-type) behaviour in 'exit choices' is invariably detrimental to evacuation systems, here, we observed that when it comes to 'adapting exit choices', a moderate degree of imitation (or followthe-peer) tendency makes the system more efficient (question II).