Understanding human decision making is a key requirement to improve crowd simulation models so that they can better mimic real human behavior. It is often difficult to study human decision making during dangerous situations because of the complexity of the scenarios and situations to be simulated. Immersive virtual reality offers the possibility to carry out such experiments without exposing participants to real danger. In the real world, it has often been observed that people tend to follow others in certain situations (e.g., unfamiliar environments or stressful situations). In this paper, we study human following behavior when it comes to exit choice during an evacuation of a train station. We have carried out immersive VR experiments under different levels of stress (alarm only or alarm plus fire), and we have observed how humans consistently tend to follow the crowd regardless of the levels of stress. Our results show that decision making is strongly influenced by the behavior of the virtual crowd: the more virtual people running, the more likely are participants to simply follow others. The results of this work could improve behavior simulation models during crowd evacuation, and thus build more plausible scenarios for training firefighters.

了解人的决策是改进人群仿真模型的关键要求，以便他们可以更好地模仿真实的人类行为。由于场景和要模拟的场景的复杂性，在困难情况下通常很难研究人类的决策。沉浸式虚拟现实提供了进行此类实验的可能性，而不会使参与者面临真正的危险。在现实世界中，经常观察到人们倾向于在某些情况下（例如，陌生的环境或压力大的情况）跟随他人。在本文中，我们研究了在撤离火车站时退出选择的人类跟随行为。我们在不同压力水平下（仅警报或警报加火灾）进行了沉浸式VR实验，而且我们已经观察到，无论压力多大，人类如何始终如一地跟随人群。我们的结果表明，决策制定受虚拟人群行为的强烈影响：虚拟人群越奔跑，参与者越容易跟随他人。这项工作的结果可以改善人群疏散过程中的行为模拟模型，从而为训练消防员建立更合理的场景。