Simulating dynamics of adaptive exit-choice changing in crowd evacuations: Model implementation and behavioural interpretations
Graphical abstract
Introduction
The fact that decision-making behaviour of humans in emergency escape or evacuation scenarios in crowded places is dynamic is well known. However, the characteristics of this dynamic mechanism is not quite well understood in detail. When evacuations happen in crowded spaces, pedestrians make various types of decisions to escape in shortest possible time (Gwynne et al., 2017). A multi-dimensional process of decision-making is a shared characteristic between evacuations of indoor spaces and those of regional and urban areas as highlighted by researchers in both domains (Kobes et al., 2010b, Mesa-Arango et al., 2012, Murray-Tuite and Wolshon, 2013, Pel et al., 2011, Pel et al., 2012). Evacuees in confined crowded spaces often have the possibility to dynamically revisit and update their decisions in response to the changes in their surroundings to ensure that they take the “best” possible strategy given the information available to them. The dynamics of the environment (e.g. the formation and propagation of the congestion) change over time during emergency evacuations and humans do have the ability and motivation to adjust their initial choices in response to those changes in order to improve their strategies. Strategy adaptation can be observed in relation to various aspects of evacuees’ behaviour including local navigation (Antonini et al., 2006a, Heliövaara et al., 2012a), global pathfinding (Kneidl et al., 2013), reaction decisions (Bode and Codling, 2018, Galea et al., 2017), exit direction choices (Bode and Codling, 2013, Duives and Mahmassani, 2012, Ehtamo et al., 2010, Fridolf et al., 2013, Haghani et al., 2015, Kinateder et al., 2018) or choice of speed (Fridolf et al., 2013).
As mentioned above, the existence of this dynamism in evacuation decision-making in complex environments is intuitively acknowledged, but for modelling purposes, it is crucial for the detailed characteristics of this behaviour to be represented in models. Particular questions that need to be taken into consideration to model this phenomenon could be as follows. What are the circumstances that may make an evacuee more likely to revise their initial decisions? What types of decisions are more likely to be updated and what are the factors driving those changes for each level of decision making (i.e. for directional exit choices, or local pathfinding choices, or global pathfinding choices). And most importantly, how can these factors be represented reasonably in numerical simulation models of evacuation (while maintaining the stability of the numerical simulation process that can potentially be affected by making decisions dynamic)?
In this work, we are particularly interested in understanding and quantifying how crucial it is to include an exit choice changing (or exit choice adaptation) feature in simulation models of pedestrian evacuations. We have observed in our past experiments that people do change their directional exit decisions. However, given that inclusion of this modelling layer will impose additional computational load and calibration complexities, we would like to quantify how important it is to model this phenomenon in the first place. Is it not good enough to approximate decisions of simulated evacuees’ as one-off decisions and generate a fixed decision for each simulated agent? What would be the impact of such modelling practice in quantitative terms?
The second central question of this study concerns the topic of behavioural optimisation. One of the end aims of modelling evacuation behaviour is to gain knowledge that can be utilised to facilitate or ‘optimise’ evacuation strategies (Kneidl et al., 2013). While broadly addressed in the context of urban network evacuations (Huibregtse et al., 2011, Murray-Tuite et al., 2012), this topic, behavioural strategy optimisation, constitutes an aspect of research in crowd dynamics that has been overlooked to great degrees (Berseth et al., 2015, Ding et al., 2017, Kou et al., 2013, Lin et al., 2008, Luh et al., 2012, Noh et al., 2016). As the main attention is currently being paid to developing descriptive models (Duives et al., 2013, Ronchi et al., 2014a) behavioural optimisation is not receiving the attention that it requires. Among the body of studies that have looked at the optimisation of evacuation processes, the majority have investigated the problem from an architectural optimisation (Luh et al., 2012, Zhao et al., 2017) or path-planning optimisation perspective (Lin et al., 2008, Noh et al., 2016, Pursals and Garzón, 2009, Vermuyten et al., 2016). In contrast, little attention has been paid to identifying optimal evacuation strategies at the level of individuals which is a major necessity for developing efficient evacuation training and guidelines (Lu et al., 2014). The question of decision changing has practical implications from the behaviour optimisation and evacuation management perspectives. In specific terms, we would like to quantitatively examine how various degrees and kinds of decision adaptation strategy influences evacuation efficiency. The question concerns whether evacuees can benefit from a dynamic decision-making strategy or is it better for them to adhere to their initial choices rather than continuously exploring and considering all available options. And if there is a benefit in choosing a dynamic strategy, how much adaptation would be optimum.
In this work, we specifically focus on adaptive decision-making in exit choices. We propose a relatively parsimonious discrete-choice model of decision changing that embodies the primary factors that influence occupants’ tendency to revise their initial exit decisions. Through full parametrisation, this model is flexibly capable of representing various forms and degrees of decision changing tendency. It also has the capacity to eliminate decision changing or activate it to extreme degrees through proper parameter specification. By implementing this model, we quantify the extent to which the inclusion of this modelling layer can improve prediction accuracy. Using this model, we also generate behavioural findings based on numerical sensitivity analyses on the parameters of this model. Utilising the fact that each model parameter carries tangible behavioural interpretations and reflects certain kind and degree of adaptation tendency, we perform extensive numerical tests to identify the types of decision-changing strategy that can benefit evacuation processes.
Section snippets
Experimental practices
In an earlier survey of the literature (Haghani and Sarvi, 2018a), we identified a substantial imbalance in the existing studies of evacuation modelling in regard to the amount of attention paid to various aspects of the evacuation decision-making behaviour. This concerns both data collection attempts as well as the development of computational methodologies for various levels of decision making. We showed that studies related to evacuees’ exit choice (i.e. tactical levels of decision making)
Model conceptualisation for exit-choice changing
The existing research on exit choice behaviour has identified a broad range of influential factors that explain such decisions including the size of queue at exits, distance to exits (Haghani and Sarvi, 2016b, Lovreglio et al., 2014), angular displacement (Duives and Mahmassani, 2012), social (or peer behaviour) influence (Haghani and Sarvi, 2017b, Kinateder et al., 2018, Kinateder et al., 2014b), exit visibility and general visibility (Cirillo and Muntean, 2013, Fridolf et al., 2013, Guo et
Comparisons with empirical observations
In order to quantify the magnitude of the effect of including a decision-change module in the simulation tool on the accuracy of the simulated modelling outputs, we resorted to two sets of experimental observations. While these experimental scenarios have been performed originally for individual-level analyses of exit choice (a type of analyses that is beyond the scope of this work), we merely used them for their macro-level (or aggregate) outputs in this work. We do not report on the analysis
How important is an exit-choice-changing module?
The outcomes of the comparison between the simulated and observed measurements have been summarised in scatterplots in Fig. 10. We contrast the observed exit shares with the simulated exit shares under the condition where the decision-change module was and was not active (plots 1 and 2, respectively). We also contrast the observed total evacuation times with the simulated total evacuation times while the decision-change module was active and while this feature was disabled (plots 3 and 4,
Conclusions
This work followed two distinct questions in relation to the evacuation behaviour and modelling of crowds. The first question concerned how crucial it is for the evacuation simulation models to represent the dynamic adaptive nature of the evacuees’ decision-making. We examined this question specifically in relation to the exit-choice behaviour. The second question looked at the effect of decision-changing mechanism on the system performance in terms of the total and individual evacuation times.
Discussion
As will be discussed in the followings, calibration of a decision-changing model (when formulated as a choice between “change” and “not change” alternatives) is relatively challenging compared to the calibration of an exit choice counterpart model. As a result, it was crucial in this work to maintain the feature of modelling parsimony in proposing the model and to limit the number of independent variables in the model. This ensures that the model remains calibratable. However, the authors would
Acknowledgements
The authors extend their sincere gratitude to the Associate Editor and three anonymous reviewers for their constructive feedback.
References (81)
- et al.
Discrete choice models of pedestrian walking behavior
Transp. Res. Part B: Methodol.
(2006) - et al.
Human exit route choice in virtual crowd evacuations
Anim. Behav.
(2013) - et al.
Dynamics of pedestrians in regions with no visibility— a lattice model without exclusion
Physica A
(2013) - et al.
State-of-the-art crowd motion simulation models
Transp. Res. Part C: Emerg. Technol.
(2013) - et al.
Movement speed and exit choice in smoke-filled rail tunnels
Fire Saf. J.
(2013) - et al.
Simulation of exit selection behavior using least effort algorithm
Transp. Res. Proc.
(2014) - et al.
Route choice in pedestrian evacuation under conditions of good and zero visibility: experimental and simulation results
Transp. Res. Part B: Methodol.
(2012) - et al.
Human exit choice in crowded built environments: investigating underlying behavioural differences between normal egress and emergency evacuations
Fire Saf. J.
(2016) - et al.
Following the crowd or avoiding it? Empirical investigation of imitative behaviour in emergency escape of human crowds
Anim. Behav.
(2017) - et al.
How perception of peer behaviour influences escape decision making: the role of individual differences
J. Environ. Psychol.
(2017)
Social dynamics in emergency evacuations: disentangling crowd’s attraction and repulsion effects
Physica A
Stated and revealed exit choices of pedestrian crowd evacuees
Transp. Res. Part B: Methodol.
Crowd behaviour and motion: empirical methods
Transp. Res. Part B: Methodol.
Hypothetical bias and decision-rule effect in modelling discrete directional choices
Transp. Res. Part A: Policy Pract.
Heterogeneity of decision strategy in collective escape of human crowds: on identifying the optimum composition
Int. J. Disaster Risk Reduct.
Imitative (herd) behaviour in direction decision-making hinders efficiency of crowd evacuation processes
Saf. Sci.
Simulating pedestrian flow through narrow exits
Phys. Lett. A
Accommodating taste heterogeneity and desired substitution pattern in exit choices of pedestrian crowd evacuees using a mixed nested logit model
J. Choice Modell.
Counterflow model for agent-based simulation of crowd dynamics
Build. Environ.
Pedestrian behavior and exit selection in evacuation of a corridor – an experimental study
Saf. Sci.
Evacuation performance of individuals in different visibility conditions
Build. Environ.
Exit choice in an emergency evacuation scenario is influenced by exit familiarity and neighbor behavior
Saf. Sci.
Social influence in a virtual tunnel fire – influence of conflicting information on evacuation behavior
Appl. Ergon.
Social influence on route choice in a virtual reality tunnel fire
Transp. Res. Part F: Traff. Psychol. Behav.
A hybrid multi-scale approach for simulation of pedestrian dynamics
Transp. Res. Part C: Emerg. Technol.
Exit choice, (pre-)movement time and (pre-)evacuation behaviour in hotel fire evacuation — behavioural analysis and validation of the use of serious gaming in experimental research
Proc. Eng.
Building safety and human behaviour in fire: a literature review
Fire Saf. J.
On the use of multi-stage time-varying quickest time approach for optimization of evacuation planning
Fire Saf. J.
A game theory based exit selection model for evacuation
Fire Saf. J.
A discrete choice model based on random utilities for exit choice in emergency evacuations
Saf. Sci.
Evacuation transportation modeling: an overview of research, development, and practice
Transp. Res. Part C: Emerg. Technol.
Effect of exit configuration on evacuation of a room without visibility
Physica A
An efficient partially dedicated strategy for evacuation of a heterogeneous population
Simul. Model. Pract. Theory
Effects of familiarity and plan complexity on wayfinding in simulated buildings
J. Environ. Psychol.
Optimal building evacuation time considering evacuation routes
Eur. J. Oper. Res.
A review of optimisation models for pedestrian evacuation and design problems
Saf. Sci.
Optimal number and location planning of evacuation signage in public space
Saf. Sci.
Optimal layout design of obstacles for panic evacuation using differential evolution
Phys. A: Stat. Mech. Appl.
Evacuation experiments in a virtual reality high-rise building: exit choice and waiting time for evacuation elevators
Fire Mater.
Discrete choice models of pedestrian walking behavior
Transp. Res., Part B, Methodol.
Cited by (52)
Human behaviour in fire: Knowledge foundation and temporal evolution
2024, Fire Safety JournalLinear and nonlinear Model Predictive Control (MPC) for regulating pedestrian flows with discrete speed instructions
2023, Physica A: Statistical Mechanics and its ApplicationsMultiple-Input-Single-Output prediction models of crowd dynamics for Model Predictive Control (MPC) of crowd evacuations
2023, Transportation Research Part C: Emerging TechnologiesAn evacuation guidance model for heterogeneous populations in large-scale pedestrian facilities with multiple exits
2023, Physica A: Statistical Mechanics and its ApplicationsBehavioral, data-driven, agent-based evacuation simulation for building safety design using machine learning and discrete choice models
2023, Advanced Engineering InformaticsCitation Excerpt :Rozo et al. [52] used a discrete choice model to capture agents’ decision-making mechanism and conducted emergency drills to calibrate the impact of distance to exits and crowd density on exit choices. Similarly, Haghani and Sarvi [53] leveraged the data from laboratory experiments to calibrate a discrete choice model that describes the influence of multiple social and environmental factors on agents’ adaptive exit-choice changes. So far, the development of ABM has been primarily based on pre-determined rules, observations, and data from surveys, laboratory experiments, and emergency drills.