The crowd dynamics under terrorist attacks revealed by simulations of three-dimensional agents

The terrorist attack has been widely modeled, in terms of crowd behavior, complex environments, attack patterns, and evacuation systems. However, most models are two-dimensional, which is unreal. The 3D factors, especially individual heights, will significantly shape both the process and outcome of terrorist attacks. Hence, the 3D model is more realistic. Taking the example of the Peshawar shooting in 2014, we apply 3D agents to reveal real-world behaviors of individuals, such as hiding, fighting, and escaping during the whole process. Based on the optimal solution out of simulations, the validity and robustness of our model can be well supported. To reveal the exact effects of key factors or mechanisms, we use counterfactual experiments and have some findings. For the fighting, more fighting of heroes can save more civilians. The hiding action of civilians can reduce the death probability. When the fighting rate in the crowd is higher, choosing to hide may be advisable for civilians. Meanwhile, hiding also has more benefits from increasing the hiding rate. A huge crowd hiding may reduce the fighting rate and then cause more deaths. Based on situations, the hiding should be chosen to maximize the gains. Our 3D model has made theoretical contributions to the field of public health and human behavior. It implies that civilians should hide properly, and be self-encouraged to be heroes, which should be advocated by the public.