Development of a Novel Group Behaviour-based Model for Pedestrian and Evacuation Dynamics


Student thesis: Doctoral Thesis

View graph of relations



Awarding Institution
Award date7 Oct 2022


Grouping is a common phenomenon in places such as streets, shopping malls, transportation systems, and others. The ‘group’ in pedestrian crowd normally covers two meanings: a ‘social group’ contains people walking together based on their social affiliations, such as families or friends; and a ‘spontaneous group’ refers to a group of people who have no social interactions but moving together displaying similar actions and outlooks due to a common goal (e.g., evacuation). The presence of these groups in pedestrian crowds can have a substantial effect on the pedestrian and evacuation dynamics but have received very little consideration in the previous studies. Nevertheless, the existing modelling researches have predominantly treated crowd as a collection of isolated individuals whereas the phenomenon of grouping is ignored. Therefore, it is necessary to account for group behaviour in the pedestrian modelling to provide more realistic evacuation simulation, which in turn contributes to crowd management and leads to sufficient but not redundant design of safer and more comfortable pedestrian facilities.

In this study, a novel group behaviour-based model is implemented on the basis of framework of the social force model (SFM) to investigate pedestrian and evacuation dynamics. We firstly formulate a new social group force that is inspired by the Lennard-Jones potential, in order to describe the dual properties of social grouping with short-range repulsion and long-range attraction. A system of model parameters is presented to describe social group characteristics and is calibrated via two controlled evacuation experiments. The good performance of the proposed model in reproducing the experimental results confirms the feasibility of our assumption on social group force. Forty-nine scenarios of room evacuation are simulated for pedestrian crowd consists of individuals and social groups with various sizes. The results show that group effects are positive for overall crowd evacuation, especially when the exit is wide. Enlarging the exit from 1 to 2 m facilitates the overall evacuation significantly, whereas the promoting effect is not substantial when the exit is larger than 2 m. An elongated configuration of the group shape is observed that orients along the moving direction and becomes increasingly notable when the leader–follower (L–F) relationship intensifies, resulting in a queue-like formation that can explain why social group behaviour facilitates the overall evacuation.

Next, the ‘spontaneous group’ with undeclared–leader–follower (ULF) structure is modelled using an information-theoretic and model-free method, transfer entropy (TE). Our proposed improvement in evacuation modelling is inspired by the real-life experience that during crowd movements, individuals usually place their visual focus on the movements of their neighbours even in the absence of social affiliations. More specifically, unlike the traditional methods describing social group with pre-assumed L–F relationship, we measure the ULF relationship in a spontaneous group according to the extended TE approach for bivariate time series given no prior knowledge. We integrate TE approach into the SFM framework to describe crowd evacuation with spontaneous group behaviour. The results show that the TE-integrated model provides more realistic trajectories than does the SFM. An individual could act as a leader and a follower simultaneously during evacuation. The ULF structure in spontaneous group is frequent at the early stage of evacuation and becomes weak when evacuees’ differences in movement states diminish.

Later, the effects of the varying TE thresholds on evacuation dynamics are examined through a series of numerical simulations. In the process of developing the TE-integrated evacuation model, we found that TE threshold is a key parameter significantly affecting the overall effectiveness of pedestrian crowd evacuation. Specifically, we visualised the dynamic characteristics of evacuees (e.g., their trajectories, movement patterns and density plot) in a room (or corridor) under various value of TE thresholds. The results show that the overall evacuation time decreases as the TE threshold increases. The trajectories occur in congested clusters when the TE threshold is low, but become more dispersed and uniformly distributed in scenarios with higher TE thresholds. The spontaneous group of larger-size (i.e., more than three followers of a leader) tends to disappear with increasing TE threshold. Lastly, we examine the location of leadership emergence. The density plot of leaders’ locations indicates that the position with the most significant influence over evacuees is directly in front of the exit and tends to move closer to the exit as the TE threshold is increased.

Lastly, we propose a TE-integrated force-based model that comprehensively considers the following behaviour and collision avoidance behaviour in pedestrian counter flows. The capability of the model is validated by comparing its performance with that of two other microscopic models for simulations of pedestrian counter flows and calibrated against a controlled experimental data set. Based on the experimental data, we summarise a new accelerate-then-decelerate lateral movement mode for pedestrians that minimises their required effort to avoid collisions with oncoming pedestrians travelling in the opposite direction. Moreover, we introduce a new parameter, ‘lane entropy’, to explain the quantitative characteristics of the lane formation phenomenon. The results show that lane entropy method outperforms the traditional order parameter method in providing meaningful evaluation of the pedestrian crowd movement order during the whole process. Our model performs well at reproducing realistic microscopic behaviours while consistently presenting the empirically observed lane formation phenomenon with quantitatively accordant self-organised features.

Finally, the conclusions of the thesis are presented, along with its limitations and appropriate suggestions for future research.

    Research areas

  • Evacuation modelling, Pedestrian dynamics, Social force model, Transfer entropy, Group behaviour, Undeclared-leader-follower structure, Collision avoidance