Designing a BIM-based serious game for fire safety evacuation simulations

Advanced Engineering Informatics 25 (2011) 600–611

Contents lists available at SciVerse ScienceDirect

Advanced Engineering Informatics journal homepage: www.elsevier.com/locate/aei

Designing a BIM-based serious game for fire safety evacuation simulations Uwe Rüppel, Kristian Schatz ⇑ Institute of Numerical Methods and Informatics in Civil Engineering, Technische Universität Darmstadt (TUD), Petersenstraße 13, 64287 Darmstadt, Germany

a r t i c l e

i n f o

Article history: Available online 30 August 2011 Keywords: Serious gaming Game design Building Information Modeling (BIM) Fire safety engineering Evacuation simulation

a b s t r a c t This paper presents results of the first phase of the research project ‘‘Serious Human Rescue Game’’ at Technische Universität Darmstadt. It presents a new serious gaming approach based on Building Information Modeling (BIM) for the exploration of the effect of building condition on human behavior during the evacuation process. In reality it is impossible to conduct rescue tests in burning buildings to study the human behavior. Therefore, the current methods of data-collecting for existing evacuation simulation models have limitations regarding the individual human factors. To overcome these limitations the research hypothesis is that the human behavior can be explored with a serious computer game: The decisions of a person during the game should be comparable to decisions during an extreme situation in the real world. To verify this hypothesis, this paper introduces a serious gaming approach for analyzing the human behavior in extreme situations. To implement a serious game, developers generally make use of 3D-modeling software to generate the game content. After this, the game logic needs to be added to the content with special software development kits for computer games. Every new game scenario has to be built manually from scratch. This is time-consuming and a great share of modeling work needs to be executed twice (e.g., 3D-modeling), at first by the architect for the parametric building model and the second time by the game designer for the 3D-game content. The key idea of the presented approach is to use the capabilities of BIM together with engineering simulations (fire, smoke) to build realistic serious game scenarios in a new and efficient way. This paper presents the first phase results of the research project mainly focusing on the conceptual design of the serious game prototype. The validation concept is also presented. The inter-operability between building information modeling applications and serious gaming platforms should allow different stakeholders to simulate building-related scenarios in a new, interactive and efficient way. Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction During the last decade, the significance of emergency management in public infrastructures has increased due to changed security conditions worldwide, which has led to the necessity of computer-aided emergency assessment process for extreme situations. Nearly every day natural as well as fire disasters or terrorist attacks are reported in the news and show the importance of making the built environment as secure as possible. Especially, in the field of Fire Safety Engineering (FSE) engineers have to face many challenges. The population growth leads to urbanization and results in mega-cities with densely populated areas. In such areas a wide spread of building types and ages are present: high-rise buildings, architectural and historical monuments, airports, railway stations and shopping malls – only to address a few – standing side by side. The decreasing number of building permits, counted by the German Federal Statistical Office (DESTATIS) [1] for both,

⇑ Corresponding author. Tel.: +49 6151 16 67 45; fax: +49 6151 16 55 52. E-mail address: [email protected] (K. Schatz). 1474-0346/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.aei.2011.08.001

new buildings and refurbishments, indicates that in Germany the building stock is slowly being renovated. For this reason, the average building age is getting increasingly old. Moreover, statistics in Germany show that the number of deaths due to fire and smoke is re-increasing since 2007 [1]. It is difficult to say whether there is a correlation between building age and death rate in case of building fire. However, according to interviews we conducted with fire fighters in the context of this research, most of these people died in older residential buildings. This fact underlines the assumption that there is deferred maintenance, particularly for fire protection, which prevents the safety level of existing buildings from being as high as it could be according to the state of the art. The challenge for civil engineers is no more to build new buildings; whereas more and more existing buildings are being refurbished and redeveloped. This new work focus is characterized by uncertainty, complexity and is conflict-ridden. During the fire safety design process, fire safety engineers have to consider a lot of prescriptive codes. This often results in the use of unnecessarily large safety factors and overly expensive costs to the building owner in case of refurbishments. For this reason, necessary refurbishments, especially to improve the safety level, run the risk of being deferred.

U. Rüppel, K. Schatz / Advanced Engineering Informatics 25 (2011) 600–611

The next section gives a brief introduction to an alternative to prescriptive codes: the performance-based fire protection design. 1.1. Performance-based fire protection design Today, fire safety engineers have an alternative to the use of prescriptive codes. Since a few years ago, a new way to reach fire safety design decisions is just being established: the performance-based approach. This process started at the beginning of the last decade driven by the ISO [2]. Many countries follow this strategy in redesigning their fire safety code systems with two parallel tracks to include performance-based as well as prescriptive regulations. In the German fire safety community a lot of work is still in progress in this particular field. With applying the performance-based approach for fire protection design decisions the focus is rather on demonstrating the safe performance of a building as a whole than meeting the detailed code requirements (e.g., height and area limits, fire-resistance ratings, egress, separations). For this purpose, it is important to understand the performance of the building and the behavior of endangered people in this building under fire exposure. During a fire event structural systems have to meet different functions (e.g., load-bearing and barrier). The load-bearing function on the one hand is important to avoid the collapse of the structural system, on the other hand the barrier function has to prevent the expansion of fire and smoke. To provide safe and smoke-free escape routes both functions must be fulfilled by the structural system. Johann et al. [3] describe an approach to integrate performance-based fire protection into the design process for structural framing systems. Johann et al. mention that it is necessary to integrate theoretical knowledge, empirical information, analytical capability and technology that has been developed by fire safety engineers into the design process. To carry out these tasks fire safety engineers make use of computer models and simulations for the description of expected spread of fire and smoke, the safety evacuation and the analysis of the overall safe performance of a building [4]. Especially, to estimate the behavior of endangered people is an essential purpose for analyzing the safety evacuation of a building, since the protection of human life is the primary aim of the performance-based approach. Beyond the people’s behavior and the corresponding human factors, the rescue mission is influenced by other factors like alarm systems, building elements and the spread of fire and smoke. In particular to map the human factors onto computer models is a challenge, because each person’s singular behavior is based on individual decisions and parameters and is not deterministic like the spread of fire and smoke, which can be modeled and simulated based on natural principles. So, according to Santos and Aguirre [5], for an evacuation simulation, three analytical dimensions need to be considered: Firstly, the built environment (physical location), secondly, the management of this environment (signage, escape routes), and thirdly, social psychological and social organizational characteristics of the occupants. Tavares [6] mentioned that an evacuation simulation model must consider four interactions: occupants–structure; occupants–occupants; occupants–fire (in case of fire events) and fire–structure (for this purpose, a fire model should be used). The next section gives a brief overview of evacuation simulation with the focus on data-collecting efforts to model the human behavior. 1.2. Evacuation simulation As mentioned in Section 1.1, considering the human factors is a big challenge for fire safety engineers in the performance-based fire protection design process. There are different reviews of existing evacuation models published in the last years [5,7] showing

601

that every modeling approach has its strengths and weaknesses. The next section focuses on the question of how to collect data for the human behavior models. Most current evacuation simulation models follow the agent-based approach. An agent is a software instance which represents a virtual person during the simulation runtime. To find out the human factors scientists use different methods like interviews, (online) questionnaires, map exercises, experimentation or the analysis of past emergency situations using interviews of survivors or CCTV recordings. The focus of the data-collecting efforts in the past was more to find out clinical mobility parameters, for example, motion speed or required space of the people according to their age or gender. These parameters were then used for statistical assumptions to describe the agent’s movement during the evacuation simulation. Nowadays, fire safety engineers are in agreement that these parameters are insufficient to describe the complex process of decision making of people during evacuations and to cover all dimensions of an evacuation simulation as mentioned in the previous section. These human factors must be considered in a more holistic manner, especially regarding the interaction with the actual building status quo. The results from the Society of Fire Protection Engineers Survey (SFPE) [8], for example, show that people often come to unexpected decisions if they are asked what they would do if they were exposed to a threatening fire. The SFPE survey concludes that some kind of behavior like fighting the fire or trying to gather belongings often results in delays before people decide to evacuate. So these delays could lead to dangerous situations, because fire and smoke spread very quickly through a building. Such behaviors are based on individual decisions and it is not quite clear why some people decide this way while others decide to evacuate immediately. Furthermore, the SFPE survey mentions that human factors, which are based on individual decisions and risk assessment, should be considered by fire safety engineers to reach better design decisions. During the research work, several software systems for evacuation simulation were evaluated. As a result, the state of the art evacuation simulation software buildingEXODUS [9], developed by the Fire-Safety Engineering Group (FSEG) at the University Greenwich, UK, was chosen for further consideration. BuildingEXODUS makes use of agent-based modeling and according to the developers it incorporates a wide range of sociological attributes and characteristics. These human factors data are collected over years as a result of different research activities in this field. Different methods, which were documented by various publications, were used for data acquisition. One example is an online survey published by Kinsey et al. [10]. In this survey, the scientists of the FSEG wanted to find out the key factors, which influence human behavior during an evacuation of a high rise building regarding the question to select the lift or the stairways. The Survey participants were presented with a series of hypothetical situations and asked how they would behave. Further within the UK High-Rise Evacuation Evaluation Database (HEED) study [11], scientists make use of interviews to collect and classify the experiences and behaviors of WTC evacuees in a database. The aim of other research methods is to improve the underlying agent-based model of buildingEXODUS by integrating wayfinding processes [12], the influence of signage [13] or to introduce a prototype emotion model into an agent based circulation simulation [14]. It is important to take all of this into account in agent-based simulations, because in reality human behavior in complex environments is dynamic and fixed plans are often changed and adjusted according to the persons’ individual interpretation of the emergent conditions. For the visualization of the simulation results of buildingEXODUS the add-on vrEXODUS is available. vrEXODUS can be used to generate 3D visualizations of the evacuation simulations in virtual reality.

602

U. Rüppel, K. Schatz / Advanced Engineering Informatics 25 (2011) 600–611

As a further result of the analysis of software systems for evacuation simulations, the agent-based simulation software FDS+Evac [15] was chosen as well for further consideration. FDS+Evac is based on the NIST Fire Dynamics Simulator [16] and uses Smokeview for 3D output. For the validation of FDS+Evac an observation of evacuation drills with video cameras and persons with Radio Frequency Identification (RFID) tags was documented [17]. The described research work above is only focused on the efforts for collecting data, which can be used to model the human behavior in evacuation simulation. Yet, a problem with the collected data is the question whether these data are resilient enough to model the complex human behavior. Santos and Aguirre [5] argue that appropriate methods for validation of the human behavior model are not available. So they state that proper validation tools have to be developed and that multidisciplinary collaboration is needed. However, according to them, there is one particular aspect that is missing from computer evacuation models: the ability to accurately and comprehensively simulate human behavior in fire [5]. Kuligowski and Gwynne [18] have taken up this criticism and suggest a new approach to model the human behavior in fire evacuation simulations. According to them, current evacuation models often have the problem that the behavior simulated in the scenario is actually prescribed by the user (with probabilistic assumptions based on collected data) rather than predicted by the model and that the current models are only simulating separated behavioral facts. They list the following behavioral facts: (1) People’s first instinct is to feel safe in their environment; (2) People will engage in information seeking actions; (3) People act rationally and altruistically; (4) People are likely to engage in preparation activities before beginning their evacuation response; and (5) People move to the familiar. But the problem pointed out by Kuligowski and Gwynne is that there is little or no understanding why these behaviors occur. Because of these facts they suggest to develop a complete, comprehensive conceptual model concerning human behavior in fire evacuations. Tavares [6] points out that difficulties occur, because fire safety engineers did not understand the phenomenon itself as a whole within all dimensions of an evacuation. However, all this shows there is need for a new approach to validate existing human behavior models. To execute this, it is necessary to find new methods for analyzing human behavior in extreme building evacuation situations. With the presented research hypothesis it is assumed that the use of serious gaming, augmented and virtual reality could offer new possibilities to solve these problems. 1.3. Virtual reality for fire safety application Advances in the simulation and automation include the use of technologies such as augmented and virtual reality for a more realistic visualization of the planned environment. In a study [19], Woksepp and Olofsson analyzed the credibility and applicability of virtual reality (VR) models for design and planning teams and for the construction site. Within their study a, 3D CAD Model, which consists of 3D objects with attributes like volume, weight and material, was manually converted into a VR model. It is documented in Ref. [19] that the differences in the definition and representation of objects in CAD and VR system currently limit a direct visualization of 3D CAD Models with VR systems. This is the result of an interoperability bottleneck when passing data between CAD and VR applications. It has been demonstrated that the respondents who use the VR models at the building site considered them to be useful and could imagine using them in their daily work particularly for the handling unfamiliar tasks. The second target group of this study were design and planning teams. In team meetings the VR models were used to examine design solutions from the different perspectives and requirements on function, work environ-

ment, and maintenance. It was documented by Woksepp and Olofsson that the VR models helped to detect clashes between the different design disciplines that minimized the risk of misinterpretation. However, one great advantage they found out was the increased understanding of the overall design and the multidisciplinary consequences of a decision. Several respondents argued that the use of VR would probably increase in future projects and that more ‘‘built-in intelligence’’ in the VR Model would extend its use in designing, planning and process simulation [19]. However, some concerns are related to the time and expenses to convert the 3D CAD Model into the VR Model and keep the VR model up to date. Trenholme and Smith [20] published the idea of using computer game technology to build virtual worlds in order to solve this problem and to minimize time effort and expenses in the complex process of building realistic virtual environments. They examined and compared different computer game engines regarding their ability for developing first-person virtual environments. Based on this idea, Smith and Trenholme explore the rapid prototyping of virtual environments with computer game technology [21]. A study conducted in this research context documented that a single developer needed around three weeks for building a realistic model of a real world building. In such a virtual building fire drill evacuation scenarios could be simulated. One result of a user study conducted in this context was that participants felt the simulated environment as realistic so that it can be assumed that virtual environments can support the training and observation of fire evacuee behaviors in 3D virtual buildings. 1.4. Summary In the performance-based fire protection design process, fire safety engineers make use of computer models and simulations for the analysis of the overall safe performance of a building. In particular, considering the human factors for evacuation simulation models is a great challenge. To examine the human factors different methods like interviews of survivors, online questionnaires, map exercises or experimentations are used. These methods often cover only singular aspects of human behavior and studies are carried out independent from each other. Another problem is that interviewees mostly know that they are not in a dangerous situation and therefore feel no cognitive emergency stress. Thus, this answering content for a questionnaire is not exactly useful to analyze human behavior in emergency situations. However, it was found out that in the FSE community there are some discussions about the point what human factors are and how they are considered in evacuation models and what proper solutions to validate the human behavior model might be. Some human factors are known, but there is only little understanding why they occur. Especially, to consider the interactions between occupants and building structure as well as the question how the destruction due to explosion or building fire influences the egress route and the corresponding decision making have only been under little examination so far. Real world experimentation cannot be conducted, due to the fact that it is simply too complex, expensive and dangerous for the participants. Virtual reality technologies can be used for a more realistic visualization of the parametric building model and help the design and planning teams to feel more ‘‘inside’’ the building and helped them to get a better understanding about it. But inside the static VR model the interactive and dynamic aspects were missing. Hence, a singular VR-Model is not suitable for putting people inside a 3D-scenario. Computer game technology can close this gap. But creating realistic game scenarios is still time-consuming, because of the large amount of modeling work that is to be executed twice (e.g., 3D-modeling), at first by the architect for the parametric building model as part of the traditional building planning, and the second time by the

U. Rüppel, K. Schatz / Advanced Engineering Informatics 25 (2011) 600–611

game artist for the 3D-game content. This indicates that further research work is required: Firstly, to develop an environment for stimulating the ‘‘egress feeling’’, secondly, to use this environment for empirical studies to find new methods for analyzing the human behavior in extreme situations, and thirdly, to use these results to evaluate existing and - if necessary - to develop new models. The focus of this paper as a report on research work in progress is on the first research part. The research hypothesis to be examined is: Human evacuation behavior can be explored with a computer game. To test this hypothesis a serious gaming approach was chosen to research new possibilities to track and observe human behavior in a controllable virtual environment. The novelty of this approach is to make use of the capabilities of Building Information Modeling (BIM) together with engineering simulations (fire, smoke) to bring BIM and serious gaming together in the area of building safety applications. 2. Approach To provide new opportunities for analyzing human behavior in extreme situations, this paper describes the approach of a BIMbased serious gaming environment for fire safety applications. The aim of the presented research is to achieve better understanding of what actually happens during an extreme situation and how an endangered person comes to decisions. Specifically, the interaction occupant-building structure is of particular interest because this could not really be investigated by real world experimentations. To work on this question, the project ‘‘Serious Human Rescue Game’’ (SHRG) started at the Technische Universität Darmstadt in December 2009. In this project, the Institute for Numerical Methods and Informatics in Civil Engineering and the Institute of Psychology are working together to research for human factors in the evacuation process with a serious gaming approach. For this approach, the building and its occupants are seen as a complex socio-technical system because there are interactions between the building (technical aspects) and human behavior (social aspects) which influence each other in a building fire scenario. The technical aspects like the spread of fire and smoke or the structural behavior of the building can be modeled and simulated based on natural principles and are more or less deterministic. The realistic behavior of endangered people - the social aspects - is more difficult to simulate and is based on individual decisions. For this reason ‘‘real’’ people have to slip into the role of these endangered people instead of software agents: but this is not possible in a real environment (a physical burning building with structural damage). Therefore, to realize this approach, the challenge is to develop a realistic and valid serious game for a new kind of immersive, dynamic and interactive simulation of building disasters. The assumed benefit of using a serious game is to track and observe human behavior in a controllable environment. 2.1. Serious gaming The approach of serious gaming combines fun methods and concepts as well as game technology with other information and communication technologies (e.g., sensors, computer graphics, multimedia, artificial technology) and sciences (e.g., computer science, design, psychology, pedagogy) in ‘‘serious fields of applications’’, beyond the pure entertainment use. Serious games are designed for the purpose of solving a task, in this case for the analysis of the human behavior during the evacuation process. Beyond the gaming approach of Trenholme and Smith described in the previous section [19,20], the market provides other games for fire safety application, mainly to train tactic and strategic skills of firefighters. With these games emergency response teams can

603

prepare themselves for incidents in a virtual environment. Many incidents can be virtually simulated in dynamic training sessions. Examples for virtual training systems are RescueSim [23] or Tactical Command Trainer [22]. The focus of these games is on the training and teaching aspect. However, no serious game that fulfills the requirements could be identified. Firstly, to provide interfaces for rapid development of scenarios based on the digital building model, and secondly, to collect the human behavior data in a satisfying manner, thus, a new serious game is to be developed. In the next step international research results around serious gaming have to be analyzed to find guidelines for designing such a game. 2.2. Triadic game design The design approach of the SHRG follows the approach of Triadic Game Design (TGD) introduced by Harteveld [24]. This approach specifically focuses on the development of computer games with serious purpose. According to Harteveld, the approach of Triadic Game Design involves a triad consisting of the interdependent worlds of reality, meaning, and play that has to be balanced out during the design process. In Fig. 1 these main components are presented with explaining information. Harteveld’s approach of TGD is based on the two design paradigm ‘‘Concurrent Design’’ and ‘‘Iterative Design’’. This means that firstly, in order to find an optimum the three worlds should be considered at the same time within critical parts of the design process (Concurrent Design), and secondly, a repeated cycle of prototyping, testing, evaluating, and redesigning continues until the requirements for the three worlds are met (Iterative Design). 2.2.1. World of reality This world deals with the question how the game is connected to the physical world. This model depends highly on the type of game and how elaborate, realistic, and valid this model needs to be. Therefore, domain-specific knowledge from the thematic background of the game is required. The challenges for developing the model of reality for the SHRG are firstly, to model this based on parametric building objects (geometry, structure and further technical semantic out of real building designs), and secondly, to enhance the model with an authentic simulation of the emergency scenario (e.g., fire, smoke, explosion). To make sure that the model is valid – underlining the ‘‘serious’’ of serious gaming – it is essential that the simulations are comparable to state of the art fire safety engineering simulations. This must be considered while creating the world of reality for the SHRG. At this point in the design process domain-specific knowledge especially from the field of civil and fire safety engineering is required. One possibility is to retrieve this knowledge from the Building Information Modeling (BIM) process by using it as a ‘‘knowledge repository’’. Meadati and Irizarry introduced how BIM can be used as a knowledge repository for a learning environment [25]. So it is assumed that BIM can also be used as a knowledge repository for a gaming environment. During the BIM process the focus is on data management of a semantical product model of a building. Building elements are represented as 3D objects with additional semantical information. BIM software tools can provide an easy access to the information, visualization, and simulation capabilities of the digital building model. Special information regarding fire protection like the ‘‘Building Information Model-Fire Protection’’ (BIM.FIRE) developed at the IIB [26,27] could be integrated in this process. 2.2.2. World of meaning The world of meaning focuses on the type of value that needs to be achieved. This is to be considered from a whole set of different disciplines and criteria. Harteveld [24] uses a frequently applied

604

U. Rüppel, K. Schatz / Advanced Engineering Informatics 25 (2011) 600–611

Fig. 1. Triadic game design according to Hartveld [24].

distinction in knowledge, skills, attitudes, and assessment, datacollection, exploration, and theory testing. So the question was: What are the values the SHRG needs to have? These values can be subdivided in values for the player and values for the observer/developer. The player should be able to acquire some knowledge about the emergency scenarios and how elements interact with each other. The SHRG can be seen as a method or tool to acquire knowledge about emergency scenarios. To achieve this value a valid model (see Section 2.2.1) for emergency scenarios is required. By immersing into this emergency scenario players achieve understanding of what happens during that moment. Another value the player should achieve is to train skills which helps to survive a catastrophe by applying the knowledge about emergency scenarios. An example here is to connect real equipment like a fire extinguisher to the serious game environment as a Human Interface Device (HID). Another aspect that has to be considered in the world of meaning is a person’s predisposition to respond in situations of the emergency scenario. These responses are often based on feelings, emotions and individual experience. Especially in emergency scenarios these feelings (e.g., fear, panic) have to be taken into account. Another value can be seen as more useful for the observer/developer of the game. Such a serious game can provide possibilities for assessment, data-collection, exploration, and theory-testing. A safe environment like a serious game could be an appropriate place to evaluate the behavior of people during emergency situation. If the structural damage could be properly simulated in the game scenario, it could be possible to assess the player’s reaction when the chosen egress route is blocked by debris or locked doors. While playing such a game, data could be collected (1) about the viewing direction, (2) at which virtual objects the player looks, and (3) in which directions he decides to move. It is

also possible to record biometrical data of the player like heart rate or brain activity at the same time. Directly after the game session additional data could be collected by a questionnaire. This collected data could be useful for achieving better understanding of how the structural status quo influences the decision making process. For exploration it should be possible to change some aspects in the emergency scenario. It is assumed that by comparing the different sessions with similar and different conditions, special aspects of an emergency scenario could be explored. Related to exploration is theory testing, for example, whether the modeled evacuation design motions based on theories, assumptions, and empirical studies are working well. The game becomes an experiment. So it should be possible for the researchers to be able to ‘‘play’’ with variables in the SHRG. Some supposable examples for these variables are the influence of alarm-sounds and emergency lightings. 2.2.3. World of play One important thing to keep in mind for the design concept of the SHRG is the type of game and in which genres it can be classified. Harteveld lists seven genres: action, adventure, puzzle, roleplaying, simulation, strategy, and virtual world games. The concept of the SHRG follows a mix of action and simulation game. According to Hartveld, the characteristics of an action game, for example, are that they do not have a long learning curve, they are fast-paced and the amount of the avatars life-energy is limited. A simulation game needs to stay close to reality and no predefined story exists (the online simulation results define the story). In the following step aspects regarding the game concept have to be discussed. For the game concept the question what objectives players have to achieve, and with what gameplay a player can reach these

U. Rüppel, K. Schatz / Advanced Engineering Informatics 25 (2011) 600–611

objectives needs to be considered. The following will take up a fire scenario as one possible scenario of a building disaster. In this virtual fire scenario a gamer plays an avatar with a special amount of vitality. The main objective for the gamer is to guide his avatar through the virtual world to a secure area before his vitality decreases until zero. Optional objectives could be that he safes as much vitality as possible or that he has to find the quickest way to a secure area. So, one requirement is that the information which influences the life-energy and the sense of orientation of a person in case of fire (e.g., heat, smoke, toxic gas), could be provided in the fire scenario. Another requirement is that it should be possible for the avatar to move through the virtual building and handle objects in the same way as in a real building (e.g., pass doors if they are not

605

locked or handle virtual equipment). The following section presents the use-cases assigned to the three worlds based on the ideas and thoughts in this section. 2.3. SHRG design concept During the design process, designers have to be equally concerned about these three worlds to balance out the game. Following the TGD approach, which is creating a model of reality, value proposal, and game concept, it is useful to know what use-cases occur, which people are involved and how they can be mixed and put together to get a balanced game. These use-cases are shown in Fig. 2 regarding the three worlds and their relevance

Fig. 2. Use-cases assigned to the three worlds (following Ref. [24]).

Fig. 3. Iterative design paradigm (according to Ref. [28]).

606

U. Rüppel, K. Schatz / Advanced Engineering Informatics 25 (2011) 600–611

for the SHRG design concept. As presented in the figure the involved persons and their activities within the concept of the three worlds are described. The defined use-cases help to figure out the concurrent parts of the SHRG concept. Now the focus is rather on the iterative part of the approach. According to Schreiber [28] several iterative steps need to be conducted. These steps are presented in Fig. 3. It starts with making an observation by pointing out the problem and developing a hypothesis (planning). Then the requirements have to be described. As a next step, existing models, methods and tools have to be analyzed and – if necessary – new models, methods and tools have to be designed. Then the implementation phase of the gaming software environment follows. When the software environment is working then it is necessary to create experiments to prove or disprove the hypothesis. Next step ahead is the execution of the experiments (testing) and the interpretation of the results (evaluation). If the evaluation is not satisfying, the next cycle starts. Otherwise the serious gaming environment is completed. The current stage of the presented research project is executed with planning, requirements, analysis, and design is up to the implementation stage during the first cycle. Thus, as a result of the research work up to now the technical aspects in developing the game engine for the SHRG are presented in Section 3. Section 4 introduces the validation concept, planned to test and evaluate the SHRG. 3. BIM-based serious gaming environment Based on the concept described in the previous section, this section gives a brief overview of the current state of development. It starts with the description of the hardware system, the Darmstadt CES-Lab, followed by an introduction to the software system, the BIM-game engine.

3.1. Darmstadt CES-lab A suitable hardware for the serious gaming environment is a virtual reality lab like the Darmstadt Civil, Environmental and Safety Engineering Lab (CES-Lab), which is established at the Institute of Numerical Methods and Informatics in Civil Engineering at Technische Universität Darmstadt (IIB). The Darmstadt CES-Lab comes with an efficient virtual environment in the sense of an immersive system. It is assumed that the immersion experience provided by this system will improve the presence of the gamer inside the computer game. Generally, virtual environments vary greatly in the quality of representing the real world. It is plausible to assume that the more accurate and richly detailed the real-world is mapped in the virtual-world as well as senses can be stimulated, the more the immersion effect in a virtual environment will increase. To test these assumptions, scientists form the Institute of Psychology are currently conducting several studies regarding active stereo capabilities, sound effects and human interface devices (HID) in the context of the SHRG project to find optimal hardware settings. A selected number of the existing and planned features are shown in Fig. 4: The main sense areas to be considered are hearing, touching, smell, tasting, and sight. In order to approximate the visual stimulus conditions, a visualization technique is used which allows the display of stereoscopic 3D-information. In order to minimize unrealistic behavior in the virtual world, the gamer should have the feeling to actually be physically inside the virtual world. Playing computer-games, gamers often tend to make unrealistic decisions due to lack of physical pain and injury responses from the virtual world. This is known in military training as the ‘‘super-soldier syndrome’’ [29,30]. The use of heating jackets, radiators, treadmills, and other devices intended to ensure that gamers do not consider themselves to be invulnerable ‘‘super-evacuees’’. Beyond that, the realism of a

Fig. 4. Existing and planned features of the Darmstadt CES-Lab.

U. Rüppel, K. Schatz / Advanced Engineering Informatics 25 (2011) 600–611

virtual environment is also influenced by the design of the audio component (surround sound) and sound effects (e.g., noise from the roaring fire, collapsing structures, crying children). 3.2. BIM-game engine The software system behind the Darmstadt CES-Lab is the BIMgame engine, which is being developed at the IIB. The BIM-game engine is comparable to a usual ‘‘game engine’’ with the ability to process BIM-data. According to Gregory [31], a game engine is ‘‘arguably a data driven architecture is what differentiates a game engine from a piece of software that is a game but not an engine. When a game contains hard-coded logic or game rules or employ special-case code to render specific types of game objects, it becomes difficult or impossible to reuse that software to make a different game. We should probably reserve the term ‘‘game engine’’ for software that is extensible and can be used as the foundation for many different games without major modification.’’ In general, a game engine is modular and consists of several components [31]. A graphics engine is responsible for the graphical display on the screen and provides interfaces for loading, managing, displaying and animating textured 3D models. One example for a graphics-engine is Ogre3D [32]. The main parts of Ogre3D are as shown in Fig. 5: resource management, scene management and rendering. Resource management is responsible to load the game objects from external resources,

607

scene management to manage the game objects during a game session and rendering to generate a picture of the scene for screen output. Beyond these main components, plugins can be added to extend Ogre3D. A physics engine, like Nvidia-PhysX [33], can simulate the mechanics of rigid-bodies and soft-bodies (building elements) or particles (smoke, fluids). A sound system provides the ability to generate realistic (surround-) sound. To save and load scenarios a game engine has to provide a data storage management. The interaction is usually executed through input devices such as keyboard, mouse, joystick and motion or gesture tracking. For gesture tracking recent developments from consumer electronics market like Microsoft Kinect [34] or Asus Wavi Xtion [35] offer new possibilities. More interactive components are network support for multiplayer mode and tools for programming and scripting at runtime without having to restart the game. The next section provides a deeper look into the BIM-game engine, the physics engine and how this can be used to simulate structural damage, fire and smoke and how the BIM-integration is implemented. In particular, the possibility to simulate the structural damage is of special interest for the research of human factors which results from the interaction occupant-structure. 3.2.1. Physics engine A physics engine is a software component that is used to simulate the dynamic physical interaction between objects on the basis

Fig. 5. Ogre3D – graphic-engine with physics and sound plugins.

608

U. Rüppel, K. Schatz / Advanced Engineering Informatics 25 (2011) 600–611

of laws of nature such as gravity and their effect on objects in virtual reality. To be able to calculate the necessary simulations in real time, strong simplifications have to be applied in favor of performance and no entirely correct physics calculations conducted, since simulation results mainly have to ‘‘look realistic’’. The simulation is normally based on the physics of rigid bodies (rigid body mechanics), on the physics of elastic bodies (soft body dynamics), and mass-spring models (spring dynamics), on which it is possible to simulate the behavior of ropes and fabrics. Modern physics engines also support particle systems, with which liquids or smoke (fluid dynamics) can be simulated; also volumetric force fields, e.g., for the simulation of heat induced flows in case of fire, can be modeled. Connections (joints) between bodies can be defined with up to six degrees of freedom. The simulation of the interaction between objects is based on collision detection and dynamics simulation. These two core components trigger certain reactions on the objects. In case of rigid body mechanics, they are mainly used to model physical reactions such as bouncing, rolling or sliding. For each 3D object the game engine requires two different representations: Firstly, a collision representation for the physics engine, in order to take the object into account during the physical simulation runtime, and secondly, a visual representation for the graphics engine for rendering the object on the screen. In comparison with the visual representation, the collision representation generally contains a simplified geometry to reduce computing time [31]. For this simplification convex shape is commonly used. Convex in the mathematical sense means that all points of an object, which lie between any two boundary points, are located inside the body. According to the number of objects, the physics engine has to perform thousands of parallel computations per simulation step. Graphics Processing Units (GPU) often come with an optimization for the calculation of many parallel tasks. Therefore, a direct hardware support for GPUs is integrated in most of the physics engines. The leading manufacturers of video cards, like Nvidia or AMD/ATI have already integrated physics support to the current graphics cards. Nvidia, for example, integrated the PhysX [33] engine (formerly NovodeX) in the CUDA system and offers a programming interface with the PhysX SDK. The main focus of the engine is the highly convenient support for Nvidia GPUs, which allows a simultaneous interaction of many objects. The SDK and part of the source code are available free of charge [33]. To synchronize the physics simulation with the graphic engine the tool NxOgre [36] is used. NxOgre is a wrapper around the PhysX SDK for the open source graphics engine Ogre3D [32]. With NxOgre for simulated objects collision, both the collision representation and the visual representation can be instantiated in virtual reality. The next section will describe the approach to approximate structural behavior of a building with PhysX for a real-time simulation during the game-runtime.

3.2.2. Approximation of structural behavior with PhysX For modeling the load-bearing system in PhysX it is important to consider firstly, the material behavior, and secondly, the joints between structural elements. This information can be retrieved from the digital building model (see Section 3.2.3). For this purpose, a structural element has to be split down into a finite number of smaller parts. For the connections of adjacent parts, transition conditions are defined, depending on whether it is subject to connections within an element (depending on the material properties like elasticity, density, and cohesion associated with the geometric definitions of the parametric BIM objects), or between structural elements (depending on bearing conditions). For the split-up the GPC polygon library is used [37].

To connect the produced smaller parts together with joints during the simulation runtime, the geometrical information is represented in a class structure. Fig. 6 shows how the subdivided parts of the parametric BIM object (in this case a wall) are connected to the class structure. CuboidStructure includes all parts of a structural element, which represents a structural element like a wall, a column or a beam and extends their parametric BIM object. A CuboidStructure is composed of several Cuboids, which in turn have several CuboidFaces to describe the faces of a body. These classes contain the necessary methods to establish connections between them. The information of a connection is stored in the class JointOptions, the material properties such as density, connection strength and textures in the class MaterialProperties. Between two Cuboids based on JointOptions and MaterialProperties, a fixed joint, a revolute joint or a triple combination (three times TwoAxesFixed) each with two limited degrees of freedom could be defined. The capabilities of these joints are used to model the material behavior within a structural element and to model the bearing conditions of the connection between two CuboidStructures. After this conversion is conducted throughout all the main elements of a building, the result is a simplified structural model for the qualitative simulation of structural behavior with the physics engine. With this real-time simulation of structural damage it is ensured that it has an influence on the gameplay, e.g., by blocking a potential egress route. 3.2.3. BIM integration As described in the previous section, the core part of the BIMgame engine is a simulation model which will include fire, smoke, explosions and structural damage. Therefore, it needs comprehensive information about a building. Requirements here are firstly, to support an international accepted data format, secondly, to have

Fig. 6. Split-up of structural elements.

U. Rüppel, K. Schatz / Advanced Engineering Informatics 25 (2011) 600–611

609

Fig. 7. File-based interchange of 3D-objects.

the semantic building information available, and finally, to have access to facility management data. As discussed in the previous section, this information should be provided via BIM. BIM nowadays is the standard modeling method for building design, so that with the SHRG a large number of real buildings could be used for the gaming environment with relatively low modeling overhead. The information coming from BIM needs to be in a format, which can be used for (1) generating the objects used by the graphic-engine for visualization, (2) by the physics engine, and (3) to setup fire and smoke simulations. The prototype of the presented BIM-game engine implements two interfaces for BIM-data. The first interface provides support for Industry Foundation Classes (IFC) [38]. IFC is an exchange format for building models. It has been specified by the IAI (International Alliance for Interoperability) and has become an international standard and is supported by most CAD-software products of the leading CAD vendors, e.g., Autodesk, Graphisoft, and Bentley. It defines an exchange format and contains object classes for storeys, roofs, walls, stairs, etc. With the IFC-format all BIMinformation is available to create the virtual environment where the emergency scenario takes place. The second interface uses Autodesk Revit API [39] to process the BIM data. Revit is a BIM software by Autodesk for 3D modeling of buildings. The modeling is accomplished in associative 3D models only. It supports parametric semantic elements such as stairs, doors or furniture. In Revit the elements are managed in an internal format and can be shared among multiple users. All information is crosslinked, so that any adjustment is automatically followed up to all affected areas. Revit supports common interchange formats like gbXML, IFC, FBX and various CAD formats. Access to the objects of the internal database is provided over the Revit application programming interface (API). The data exchange between BIM-Software (here Revit) and the BIM-game engine is implemented via

file system. Fig. 7 shows the process of data-exchange. The component RevitToOgre can generate the 3D-Game-Elements for Ogre3D (.mesh) and PhysX (.nxs). To re-create the building in the BIMgame engine the information of positions and constraints (semantical data) of the building parts are required. This information is stored in separate files (.obim, .opos, .ojnt). The component BIMGameStarter reads the information stored in the different files and puts them together according to the game scenario. The component BIMGame is the simulation-core and is responsible for controlling the gameplay.

4. SHRG validation concept One major challenge in following this serious game approach is the question of validity. Does the SHRG replicate real building fire scenarios in a satisfying way? For validation of the SHRG a comparable real environment has to be found. For such a real environment a BIM-model has to be created and used as a game scenario basis. An adequate real world environment was found in the Fire House of the fire brigade academy of the state of BadenWürttemberg [40]. Fire-fighters deal with fire in their daily work and have broad experience with it. They have to train their operations regularly to be ready for emergency situations. Real-life training in fire houses is the commonly most acknowledged way to develop competences in the extremely important hot part of the firefighters’ education, which is being face to face with the flames. This is one of the significantly involved prerequisites to ensure that firefighters are able to complete their mission successfully. They need to develop a feeling for the conditions in a burning building – roaring flames, temperatures of more than 700 °C and toxic smoke can have serious consequences even if the smallest of mistakes occurs under the high stress for body and mind during an

610

U. Rüppel, K. Schatz / Advanced Engineering Informatics 25 (2011) 600–611

Fig. 8. Fire house of the state Baden-Württemberg in Germany.

Fig. 9. Validation concept of the SHRG.

operation. To learn about the effects of smoke, heat, fire extinguishing, and evaporating water, the firefighters can train in fire houses like the one of the fire brigade academy of the state Baden-Württemberg (see Ref. [40]). Fig. 8 displays the Fire House from outside and one firefighter in action. As it is shown in Fig. 8, a fire house is a real building with rooms, each of them differently equipped for simulating realistic scenarios for firefighting. The computer-controlled burning unit is driven with gas and different scenarios like a kitchen-, a warehousing-fire or a ‘‘flashover’’ can be simulated. However, the idea to evaluate SHRG is to implement the fire house described above as one gaming scenario. The workflow is shown in Fig. 9. Therefore, a BIMmodel of the real world environment is still under development in cooperation with LFS-BW: This will be the basis for the serious game scenario. With this, firefighters should be able to play through the same scenarios both in the real Fire House and in the SHRG. After that, the firefighters should be asked for their individual impression on reality in the SHRG. This would provide potentially useful clues for further development and reviews for the SHRG’s quality. 5. Future work After the game prototype development is finished in a first step, a validation experiment following the concept described in

Section 4 is planned. This helps to figure out how to adjust the virtual environment to stimulate the desired effect. Concurrently, it is planned to enhance the underlying structural model for the fire and smoke simulation, so that the results become more physically realistic instead of only ‘‘visually satisfying’’. In order to realize the follow-up step, there are plans to use the optimized SHRG to research human factors during extreme situations. If the validation would prove the developed game to be a valid tool for analyzing human behavior, many benefits might be possible. Then, the SHRG could be used for data collection, exploration and theory testing in the field of evacuation under building fires. Another future opportunity could be seen in using the collected human factors to train machine-based learning environments. These environments could be used for an improved prediction of the agent’s behavior in agent-based evacuation simulation models. This approach, for example, is used for robots to be able to predict pedestrians’ movements [41] based on Markov decision processes (MDPs) [42]. However, eventually, the system developed according to the approach described in this article, should enable new analysis methods for studying the human behavior in extreme situations within the performance-based fire safety design and it is assumed that particularly, the human factors regarding the interaction occupant-building structure could be explored in a new way. This could help to improve existing or develop new evacuation models.

U. Rüppel, K. Schatz / Advanced Engineering Informatics 25 (2011) 600–611

6. Conclusion This article introduces the concept of a new kind of a BIM-based serious game, the Serious Human Rescue Game, for an interactive and real-time simulation of emergency scenarios using a serious gaming approach. According to the concept, BIM-models serve as a basis of the game scenario. This allows a quick setup of scenarios by improving the interoperability to pass the data between the BIM application and the serious game. This feature provides a close relation to the world of reality (BIM-models are the standard data pool for building facility management) and allows designers, creators, operators, and users of the building as well as rescue forces to simulate various scenarios within the shortest amount of time without extra modeling of the game scenario. It is assumed that these capabilities can help to avoid clashes and to detect security risks in an early planning stage. With the integrated physics engine qualitative simulations of fire and smoke as well as structural damage after explosions or earthquakes are possible.

[11] [12]

[13]

[14]

[15] [16]

[17]

[18]

7. Future directions and challenges [19]

The introduced serious game is a new and efficient implementation method for bringing BIM and serious gaming together in the area of building safety applications. This combination has the potential to set a new technology trend in using BIM for all serious games with a close relation to the real built world. BIM software tools can provide the bridging technology between the real world and the virtual game world. But currently, as one result of the project, not all information, which is required for generating realistic game scenarios, is included in BIM. Additional information needs to be retrieved from different resources or converters. For example, major BIM tools provide functionality and integrating capabilities for 3D data interchange technologies and facilitate data exchange between several content creation tools for creating game objects. But the definition of dynamic game objects for a realistic and immersive real time simulation of building related game scenarios is still associated with a major effort. The future direction and challenge is to provide BIM-based serious gaming environments in such a way that they become a common engineering method for daily practice with new possibilities for the analysis and exploration of a wide range of building related aspects. References [1] DESTATIS GENESIS-Online, Statistisches Bundesamt Deutschland, 2011. (accessed May 2011). [2] ISO TC 92, Res.244 TG N6 Framework for the long term standardization of fire safety in support of performance-based design, 2002. [3] M.A. Johann et al., Performance-based structural fire safety, Journal of Performance of Constructed Facilities ASCE (February) (2006). [4] Raymond Friedman, An international survey of computer models for fire and smoke, SFPE Journal of Fire Protection Engineering 4 (3) (1992) 81–92. (accessed May 2011). [5] G. Santos, B.E. Aguirre, A critical review of emergency evacuation simulation models, in: NIST Workshop on Building Occupant Movement during Fire Emergencies, 2004. [6] R.M. Tavares, Evacuation Processes Versus Evacuation Models: ‘‘Quo Vadimus’’? Fire Technology, 2008, ISSN:1572-8099. [7] E.D. Kuligowski, R.D. Peacock, A Review of Building Evacuation Models, NIST Technical Note 1471, National Institute of Standards and Technology, 2005. [8] C. Jelenewicz, Human perceptions and reactions to fire: results from society of fire protection engineers annual survey, Journal of Architectural Engineering ASCE (March) (2008). [9] FSEG, Exodus Introduction, Information, 2011. http://fseg.gre.ac.uk/exodus/ index.html> (accessed May 2011. [10] N.J. Kinsey, E.R. Galea, P.J. Lawrence, Stairs or Lifts? – A Study of Human Factors associated with Lift/Elevator usage during Evacuations using an online Survey

[20] [21] [22]

[23] [24] [25]

[26]

[27]

[28] [29]

[30]

[31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42]

611

2010. (accessed September 2010. N.C. Mcconnel et al., The UK9/11evacuationstudy: Analysisofsurvivors’ recognitionand responsephaseinWTC1, Fire Safety Journal 45 (2010). A. Veeraswamy, P.J. Lawrence, E.R. Galea, Implementation of cognitive mapping, spatial representation and wayfinding behaviours of people within evacuation modelling tools, in: Proceedings of the 4t International Symposium on Human Behaviour in Fire 2009 Interscience Communications, 2009, ISBN:978-0-9556548-3-12009. H. Xie, L. Filippidis, E.R. Galea, D. Blackshields, P.J. Lawrence, Experimental study of the effectiveness of emergency signage, in: Proceedings of the 4t International Symposium on Human Behaviour in Fire 2009 Interscience Communications, 2009, ISBN:978-0-9556548-3-12009. C. Hollman, P.J. Lawrence, E.R. Galea, Introducing Emotion Modeling to AgentBased Pedestrian Circulation Simulation, 2010. (accessed May 2011). VTT, Fire Dynamics Simulator with Evacuation (FDS + Evac), 2011. (accessed May 2011). NIST, Fire Dynamics Simulator and Smokeview (FDS-SMV,) National Institute of Standards and Technology (NIST), 2011. (accessed May 2011). VTT, VTT Working Papers 85: Experimental observations of evacuation situations, 2007. (accessed May 2011). E.D. Kuligowski, S. Gwynne, The need for behavioral theory in evacuation modeling, in: Pedestrian and Evacuation Dynamics 2008, Springer, Berlin, Heidelberg, 2010. ISBN:978-3-642-04504-2. S. Woksepp, T. Olofsson, Credibility and applicability of virtual reality models in design and construction, Advanced Engineering Informatics 22 (2008). D. Trenholme, S.P. Smith, Computer game engines for developing first-person virtual environments, Virtual Reality 0 (2008) 181–187. S.P. Smith, D. Trenholme, Rapid prototyping a virtual fire drill environment using computer game technology, Fire Safety Journal 44 (4) (2009) 559–569. Vectorcommand Ltd., Tactical Command Trainer, Product Information, 2011. (accessed May 2011). VSTEP, RescueSim-Virtual Emergency Response Training, Product Information, 2011. (accessed May 2011). C. Harteveld, Triadic Game Design – Balancing Reality, Meaning and Play, Springer, London, 2011. ISBN:978-1-84996-157-8. P. Meadati, J. Irizarry, BIM – a knowledge repository, in: Proceedings of the 46th ASC Annual International Conference, 2010. (accessed May 2011). U. Rüppel, M. Lange, A. Wagenknecht, Semantic integration of product model data in fire protection engineering, in: Proceedings of the European Conference on Product and Process Modelling (ECPPM), e-Business and e-work in Architecture, Engineering and Construction, Valencia, 2006, pp. 115–118. M. Theiss, Agentenbasierter Modellverbund am Beispiel des baulichen Brandschutzes in der Gebäudeplanung, Technischer Bericht Nr. 3/2005 des Instituts für Numerische Methoden und Informatik im Bauwesen, ShakerVerlag, Darmstadt, 2005. I. Schreiber, Game Design Concepts, Weblog, 2011. (accessed May 2011). M.Barlow, P. Morrison, Challenging the Super Soldier Syndrome in 1st Person Simulations, 2005. (accessed May 2011). P. Morrison, M. Barlow, G., Bethel, S. Clothier, Proficient soldier to skilled gamer: training for COTS, 2005. (accessed May 2011). J. Gregory, Game Engine Architecture, Wellesley: AK Peters, 2009. OGRE3D, Object-Oriented Graphics Rendering Engine, 2011. (accessed May 2011). NVIDIA, Nvidia PhysX Technology, 2011. (accessed May 2011). Microsoft, Xbox Kinect, 2011. (accessed May 2011). ASUS, Asus Wavi Xtion, 2011. (accessed May 2011). NXOGRE, NxOgre – Physics wrapper for PhysX and Ogre3D, 2011. (accessed May 2011). A. Murta, GPC-General Polygon Clipper library, 2011. (accessed May 2011). T. Liebich, IFC 2x Edition 3 Model Implementation Guide Version 2.0 building, SMART International Modeling Support Group, 2009. Autodesk, Autodesk Revit 2010 Product, 2011. (accessed May 2011). LFS-BW, Landesfeuerwehrschule Baden-Württemberg, 2011. (accessed May 2011). B.D. Ziebart et al., Planning-based prediction for pedestrians, in: Proceedings IROS 2009, 2009. R. Bellman, A Markovian decision process, Indiana University Mathematics Journal 6 (1957) 679–684. ISSN:0022-2518.