A Simulation Study of Exit Choice based on Effective Throughput of an Exit Area in a Multi-Exit Evacuation Situation Kashif Zia Institute for Pervasive Computing Johannes Kepler University Linz, Austria kashif.zia@researchstudio.at Alois Ferscha Institute for Pervasive Computing Johannes Kepler University Linz, Austria ferscha@soft.uni-linz.ac.at Abstract—To individuals evacuating, many multi-exit environments do not allow visibility of all the exits due to line- of-sight constraint. In addition, the environment can be dark or smoky, not allowing visibility to even a single exit. In such a situation, given that each individual in the crowd is accompanied with a helping device globally connected with a central server, a 'directional guidance' towards an optimal exit is a real possibility. In this context, the 'occupant density' around exits (within a static ‘exit area’) has been used in conjunction with the corresponding distances to devise a probabilistic strategy for optimal exit suggestion [1]. In this paper, we related the exit area with the level of visibility of the environment (the more the visibility, the more the exit area and vice versa). In this way, a more realistic human-behavioral model is implemented in which an individual viewing (seeing) an exit would always direct towards that exit, irrespective of the directional guidance provided. When an individual is not at any of the exit areas (not viewing even a single exit), a directional guidance is provided assuming that the individual is adhering to it. Additionally we used the measure of 'effective throughput' instead of occupant density, in conjunction with the corresponding distances. Through simulation results, we found a marked improvement in the evacuation time, when effective throughput was modeled instead of occupant density. Crowd Evacuation, Occupant Density, Crowd Simulation. I. INTRODUCTION Most crowd evacuation simulation systems are based on human perceptions, particularly the sense of sight and hearing. In a socio-technical system, some (or more) individuals are equipped with at least one helping device, thus providing a possibility of utilizing device-to-device interaction for more efficient evacuation. Device-to-device interaction is dependent on the communication range of the devices and may result in the formation of different communication zones, for example, local neighborhood, clusters or groups (of varying sizes) and a global network. Group dynamics, being a more complex and demanding phenomena, involve more social behavioral aspects, thus resulting in a deviation from a 'simple-to-start' socio- technical system in which humans and devices have a one- to-one relation. That is the reason, in this paper; we only concentrate on the local neighborhood on one hand and a global network on the other. These effectively two extremes are more popularly categorized by the research community as microscopic and macroscopic interaction. Emerging crowd behavior at the macroscopic level cannot be modeled without understanding of behavior of the individuals at the microscopic level [2]. Alternatively, modeling behavior at the local level accumulates to an emerging behavior at the global level. Understanding the dynamics of emerging behavior at the macroscopic level, potentially suggests a continuous feedback to the individuals, helping them readjust their local behavior to achieve the desired patterns at the global level. More specifically, during the crowd evacuation process, through the microscopic level interaction, an individual (or a device holder) may know about the local neighborhood and make a decision about the next site to move to. If the macroscopic interaction is active (suggestions provided by a server to the individual), the decision about the next site would be under the influence of the exit suggestion provided. A cumulative microscopic activity in a region changes the global situation, and may result in a slightly different suggestion to the same individual for the next time. The most prominent global goal of any evacuation system is to increase the overall evacuation efficiency. Mostly, the evacuation efficiency is a measure of evacuation time i.e. the iterations required to evacuate all (or most of) the population. Given that each of the individual is equipped with a device, the environmental conditions (particularly its location) can be communicated to a central server. As a result of application of a macroscopic model at the server, each individual (or device) in the population may be provided with the optimal directional guidance. A directional guidance provided in this way is indigenously dependent on the macroscopic view of the environment based on the phenomena of interest. For the systems targeting the evacuation efficiency, the phenomena of interest is related with the structure of the environment. For example placement & level of threat(s), damaged regions & presence of obstacles, geometry of confinement area, population density and the visibility level. Population density can be defined as a number of individuals in a region. Traditionally, researchers have used Exit Area (EA), i.e. area surrounded by an exit, as a more effective region to measure the population density; the measure known as Occupant Density (OD) at an exit. In this respect, the main premise of the evacuation strategies is that an individual potentially diverts from 'nearest' exit and selects one of the remaining exits as its destination, considering the OD at exits [3]. Mathematically this strategy was introduced 2009 13th IEEE/ACM International Symposium on Distributed Simulation and Real Time Applications 1550-6525/09 $26.00 © 2009 IEEE DOI 10.1109/DS-RT.2009.13 235