Evacuating with Others Virtually

由于火警/烟雾警报器、化学品泄漏警报、炸弹威胁等，现代建筑物每天都面临紧急疏散。虽然大多数情况下是虚惊一场，大多数疏散都是有序的，但也有一些悲剧的结局，给当事人和社会带来了巨大的损失。疏散过程中的困难包括决定在有限的时间内做什么，但会产生严重的后果，了解威胁的动态性质（例如，火焰/烟雾或化学品泄漏的传播），以及找到没有过度拥挤的安全通道。贯穿这些问题的一个事实是，所有此类疏散都涉及多名撤离人员的同时移动。因此，人类在压力下的社会心理行为是至关重要的，迫切需要在受控环境下设计和进行社会心理学实验，并制定可验证的理论。这项研究将使用沉浸式虚拟现实（VR）实验——一种将个人嵌入建筑物紧急情况的足够吸引人的方式——与跑步机和生理评估一起提交科学测试假设，否则无法通过实验进行测试。此外，我们将建立数学疏散模型，考虑实验中确定的关键社会心理特征，以预测疏散人员在与其他人一起疏散时的行为。然后将开发新一代优化方法，以确定如何以计算效率高的方式有效地将撤离人员引导到安全地点。模型和方法将通过仿真、VR实验和消防演习进行部分验证。通过将一个由工程师、社会和感知心理学家以及安全问题专家组成的跨学科团队聚集在一起，该项目将建立一个经过良好测试的关于疏散过程中社会影响的原则，以建立一个疏散计划，该计划可以确定可能发生的事情，以及如何有效地引导疏散人员到安全的出口或消防避难区域。该项目还将产生更广泛的影响，为教育学生、专业人员和社区安全疏散提供独特而丰富的环境。具体来说，VR实验将在研究中包含两个部分。在第一部分中，个人参与者的实验将提供一些方式的严格测试，其中信息和社会影响过程已经假设影响疏散。在第二部分中，小群体参与者实验将提供一个适度控制的环境，以探索当疏散人员群体撤离时，某些现象会发生的条件，这些现象不可能通过单独关注个体参与者来检查，除非在实际灾难等不受控制的环境中，否则不可能观察到。要研究的关键问题包括基于信息的焦虑减少、对环境的熟悉程度（或习惯性撤离倾向）、羊群（或与他人联系的吸引力）、对指导的反应和社会纽带因素的假设。将建立数学疏散公式，以捕捉在VR实验中确定的社会心理特征的本质。然后将开发新一代随机混合整数优化方法，使用我们最新的代理拉格朗日松弛在分解和协调框架内以计算高效的方式优化指导。个别群体的出口路线将分别得到优化，由此产生的路线将得到协调，以便适当地共享容量有限的通道和房间，以满足联合行动的总体需求。模拟、虚拟现实实验和消防演习将为模型和方法提供部分验证。我们希望能够看到瓶颈的出现和消散，并检查实验操作如何影响这些和其他现象。我们还希望能够对这些现象进行建模和预测，并开发优化的解决方案，以减轻阻塞和其他有害行为，从而实现有效的疏散。

Triggered by fire/smoke alarms, chemical spill alerts, bomb threats, etc., emergency evacuations are faced daily by modern buildings. Although most of these cases turn out to be false alarms and most evacuations are orderly, some end tragically and cost greatly to those involved and to the society. Difficulties during evacuations include deciding what to do with very limited time but serious consequences, understanding the dynamic nature of threats (e.g., propagation of fire/smoke or chemical spills), and finding safe passages which are not over-crowded. Cutting across these issues is the fact that all such evacuations involve the simultaneous movement of multiple evacuees. As such, human social psychological behaviors under stress are of paramount importance, and there is a critical need to design and conduct social psychological experiments under controlled circumstances and to formulate verifiable theories. This research will use immersive virtual reality (VR) experiments - a sufficiently engaging way of embedding an individual in a building emergency - with treadmills and physiological assessments to submit to scientific test hypotheses that are otherwise impossible to test experimentally. Moreover, we will establish mathematical evacuation models taking into account the key social psychological features identified in the experiments to predict how evacuees will behave when evacuating with others. A new generation of optimization methodology will then be developed to determine how evacuees should be effectively guided to safe locations in a computationally efficient manner. Models and methods will be partially validated through simulation, VR experiments and fire drills. By bringing together a cross-disciplinary team of engineers, social and perception psychologists, and experts in security issues, the project will establish well-tested principles about social influences during evacuation to bear on creating an evacuation program that can determine what is likely to happen and how to effectively guide evacuees to safety, either exits or fire refuge areas. The project will also have the broader impact of providing a unique and fertile context for educating students, professionals and communities about safe evacuations.Specifically, VR experiments will contain two parts in the research. In the first part, individual-participant experiments will provide rigorous tests of a number of ways in which information and social influence processes have been hypothesized to affect evacuation. In the second part, small group participant experiments will offer a moderately controlled setting for exploring the conditions under which certain phenomena will occur when groups of evacuees evacuate, phenomena that are impossible to examine by a focus on individual participants alone, and impossible to otherwise watch unfold except in uncontrolled settings such as actual disasters. Key issues to be examined include hypotheses on information-based anxiety reduction, familiarity of the environment (or habitual evacuation tendencies), herding (or pull to affiliate with others), responses to guidance, and social bond factors. Mathematical evacuation formulations will be established to capture the essence of social psychological features identified in VR experiments. A new generation of stochastic mixed-integer optimization methods will then be developed to optimize guidance in a computationally efficient manner using our latest Surrogate Lagrangian Relaxation within the decomposition and coordination framework. Egress routes for individual groups will be separately optimized, and resulting routes will be coordinated so that passages and rooms with limited capacities are properly shared to meet the total needs for joint movement. Simulation, VR experiments and fire drills will then provide partial validations of models and methods. We expect to be able to see bottlenecks appearing and dissipating, and to examine how experimental manipulations affect these and other phenomena. We also expect to be able to model and predict such phenomena, and develop optimized solutions to alleviate blocking and other detrimental behaviors for effective evacuation.