DEGREE PROJECT IN COMPUTER ENGINEERING, FIRST CYCLE, 15 CREDITS STOCKHOLM, SWEDEN 2018 Evaluating classroom evacuation with crowd simulation WILLY LIU DANIEL PARHIZGAR KTH ROYAL INSTITUTE OF TECHNOLOGY SCHOOL OF ELECTRICAL ENGINEERING AND COMPUTER SCIENCE Evaluating Classroom Evacuation with Crowd Simulation WILLY LIU, DANIEL PARHIZGAR Bachelor in Computer Science Date: June 6, 2018 Supervisor: Christopher Peters Examiner: Örjan Ekberg Swedish title: Evaluering av klassrumsevakuering med simulering av folkmassor School of Electrical Engineering and Computer Science iii Abstract Effective organization of evacuation routes inside rooms needs to be considered when designing the layout. Studies have shown that an in- effective evacuation can in itself be the cause of injuries in emergency situations. This thesis examines how exit configurations and different layouts in classrooms will impact evacuation time using crowd simu- lation through unilaterally incompressible fluid. The classrooms used in the simulations were modelled after real classrooms at KTH Royal Institute of Technology. The different configurations were one exit, two exits, three exits, one exit on the back wall, and a double door on the back wall. These were chosen since previous research have focused on similar layouts. The simulator was used to simulate students evacuating from the different classrooms. The results suggests that which exit configura- tion is the most optimal is linked to the table layout of the classroom. In the simulated scenarios, two doors always result in faster evacua- tion time compared to one door. A double door in the back also de- creases evacuation time compared to one door. Three doors do not give significant improvements over two doors. Contrary to previous research, there exists no optimal exit configuration that is optimal in all cases. iv Sammanfattning Effektiv organisering av utrymningsvägar inuti rum måste beaktas vid utformningen av layouten. Studier har visat att en ineffektiv evakue- ring i sig kan orsaka skador i nödsituationer. Denna rapport under- söker hur utgångskonfigurationer och olika layouter i klassrummen kommer att påverka evakueringstiden med hjälp av folkmassimule- ring genom ensidigt inkompressibel vätska. De klassrum som använ- des i simuleringarna modellerades efter riktiga klassrum vid Kungliga Tekniska Högskolan. De olika konfigurationerna var en utgång, två ut- gångar, tre utgångar, en utgång på bakre väggen och en dubbeldörr på bakre väggen. Dessa valdes eftersom tidigare forskning har fokuserat på liknande layouts. Simulatorn användes för att simulera studenter som evakuerar från de olika klassrummen. Resultaten föreslår att vilken utgångskonfigu- ration som är optimal är kopplad till klassrummets bordslayout. I de simulerade scenarierna resulterar alltid två dörrar i snabbare evaku- eringstid jämfört med en dörr. En dubbeldörr vid baksidan minskar också evakueringstiden jämfört med en dörr. Tre utgångar ger inte märkbart bättre resultat än två dörrar. Till motsats till vad tidigare forskning har visat finns det ingen utgångskonfiguration som är op- timal i alla situationer. Contents 1 Introduction1 1.1 Research question.......................2 1.1.1 Purpose........................2 1.2 Scope..............................2 1.3 Report outline.........................3 2 Background4 2.1 Unity..............................4 2.2 Crowd simulation.......................4 2.3 Unilaterally incompressible fluid..............5 2.4 Crowd evacuation.......................7 2.4.1 Faster is slower effect.................8 2.5 Related works.........................8 2.5.1 Crowd simulation state of the art..........8 2.5.2 Evacuation simulation................9 3 Methods 12 3.1 Implementation........................ 12 3.1.1 Modeling room layout................ 12 3.1.2 Building the simulation environments....... 14 3.1.3 Modelling 3D objects................. 14 3.1.4 Running the simulations............... 14 3.2 Evaluation........................... 15 3.2.1 Collecting and evaluating the data......... 15 3.2.2 Comparing to previous research.......... 15 4 Results 16 4.1 Small classroom........................ 17 4.2 Medium classroom...................... 19 4.3 Large classroom........................ 21 v vi CONTENTS 4.4 Comparing number of exits................. 23 5 Discussion 24 5.1 Result discussion....................... 24 5.1.1 Number of exits.................... 24 5.1.2 Exit size and placement............... 25 5.1.3 Room layout...................... 26 5.1.4 Summary....................... 27 5.2 Limitations........................... 27 5.3 Reflection............................ 28 5.4 Future work.......................... 28 6 Conclusion 30 Bibliography 31 Chapter 1 Introduction Effective organization of evacuation routes inside buildings is an im- portant aspect that architects need to consider when designing a build- ing. With the rising number of emergency situations occurring in school settings, it is important to be able to evaluate a classroom, in order to have a classroom that is as optimized as possible for evacuations. However, human crowds can exhibit complex and/or unpredictable behavior during emergency situations [9]. Due to limited knowledge of human behavior, accurately predicting how people behave is diffi- cult [4]. In extreme cases of emergency situations, an ineffective evacu- ation can itself lead to injuries due to stampedes/people getting tram- pled as the crowd races to exit [5]. Although it is certainly possible to actually build the classroom to evaluate it, it might not give realistic results; The panic factor and other behaviors that arise in emergency situations are lost in drills, and ex- posing people to real danger cannot be considered a viable evaluation method. Simulating real life with computer programs can be a good way to research and draw conclusions about certain behaviours in different settings. One interesting subject within simulation is simulating crowd behaviours which can be useful when planning designs for the inside of a building, effective walkways in areas with dense crowds, or when designing evacuation plans. A crowd simulation will help reduce the cost and danger of creating and testing different classroom layouts for evacuation while still getting accurate results that can be applied in real life. 1 2 CHAPTER 1. INTRODUCTION 1.1 Research question This paper will investigate how crowd simulation can help with the evaluation of classroom evacuation times when the exit configuration and room layout is different. Specifically, we will focus on real class- room layouts based on typical classrooms present in KTH Royal Insti- tute of Technology. Our research question is the following: How will exit placement, number of exits, exit size, the size of the room, and the room layout impact emergency evacuation time in classrooms? 1.1.1 Purpose Answering our research question is important since previous research have come to different conclusions regarding this subject and will lead to better knowledge about which exit layout works best for emergency evacuation. Answering this research question will also let us know how good of an evaluation tool simulation via unilaterally incompressible fluid is. This is important because this simulation model is relatively new [11] compared to other simulation methods [6] and it is therefore good to do more research on it. 1.2 Scope Our thought process when delimiting ourselves is to aim for realistic behaviour in real life settings rather than realistic appearance, such as how realistically the people walk or the environment graphics. We will also limit ourselves to evacuation in classrooms, specifi- cally those of KTH Royal Institute of Technology which we have easy access to. We have limited ourselves to three specific classrooms of dif- ferent sizes, small, medium, and large (see chapter3 for more details). The rooms will also be filled with people and furniture according to what the blueprints of the classroom says and the max amount of peo- ple allowed in the rooms. When deciding on multiple exit placements, there are potentially limitless different variations one could come up CHAPTER 1. INTRODUCTION 3 with, but we want these to be located on the same wall to be scenarios being more realistic. However, we purposely try a layout with a door at the back of the classroom since previous research [18] specifically indicates that this is the best layout. Placing an exit on the actual back wall in the chosen classrooms is in practice not possible due to the lo- cation of said classrooms, but in real life it is possible to re-organize the furniture in a room to change where the back of it is located. The results should therefore still be applicable in real life. We also assume that the classrooms are located on floors where the people can’t use the windows as exit. Finally, another limitation we acknowledge is that real crowds have elements of randomness such as variation in people’s height, width, walking speed etc that is not taken into account in this report. We have also limited ourselves to simulations where each student’s starting po- sition is at one of the available seats in the classroom. In real life, the students could potentially be walking around in the classroom, lead- ing to a less even distribution, but we have chosen what we believe is the most representative scenario - when students are by their desks. 1.3 Report outline This report will begin by outlining all relevant theory in chapter2. Concepts such as crowd simulation and crowd