USING COMPUTER GAMING TECHNOLOGY TO EXPLORE HUMAN WAYFINDING AND NAVIGATION ABILITIES WITHIN A BUILT ENVIRONMENT
Tim Germanchis1, William Cartwright2, Christopher Pettit3
1,2 School of Mathematical and Geospatial Sciences, RMIT University, GPO Box 2476V, Melbourne, Victoria 3 Department of Primary Industries, Werribee, Victoria 1 Tel: +61 3 9925 1729, Fax: + 61 3 9663 2517, Email: [email protected] 2 Tel: +61 3 9925 2423, Fax: +61 3 9663 2517, Email: [email protected] 3 Tel: +61 3 9925 3271, Fax: +61 3 9663 2517, Email: [email protected]
ABSTRACT
Three-dimensional real-time representations of geographical data on computers are known as Geospatial Virtual Environments or GeoVEs. Previous work in the display of GeoVEs utilises a conventional multimedia/hypermedia approach. However, in recent times, large technological advances have seen computer gaming technology offer a fitting environment for serious academic study. Much literature has argued that GeoVEs lack meaningful interaction and realism, especially as compared to games. Therefore, research into applying games technology to GeoVEs may be satisfactorily used to create a more interactive, realistic and hence a more engaging environment. This paper explores the potential of utilising a games approach in displaying GeoVEs and demonstrates a ‘proof-of-concept’ prototype developed at RMIT University’s School of Mathematical and Geospatial Science, built using Crytek’s Far Cry game engine.
The intended ‘proof-of-concept’ prototype will model the town of Queenscliff, Victoria, Australia. It is a coastal town boasting historic buildings, situated on top of a slightly undulating landscape. Queenscliff was chosen because of its urban extent, scenic backdrop, and its aerial coverage is a reasonable modelling size for a gaming engine to handle.
Our research aims to further contribute to the body of research in wayfinding and navigation within virtual cities. By referencing this prototype to a real world urban planning issue, the impetus behind human spatial cognition derives from the famous urban planner, Kevin Lynch, who in his book, “The image of the city” (1960) he describes a study that looked at how people build a mental representation of the city in which they live in. Building upon his tests and ideas, it is considered that the provision of strategically placed elements will improve ‘place legibility’ of a GeoVE, allowing for improved virtual spatial movement by participants.
KEYWORDS
Games engines, Geospatial Virtual Environments (GeoVEs), Wayfinding, Navigation, Urban Planning, Usability.
INTRODUCTION
The choice of a suitable modeling environment relies on the task at hand, which is to develop a virtual representation of the existing township of Queenscliff. Visualizations of large-scale urban environments have traditionally been undertaken through the use of existing multimedia/hypermedia tools such as GeoVRML, Flash, Director3D etc., Computer Aided Design (CAD) or Geographic Information System (GIS) applications (Pettit et al. 2003). Many of these applications do not have the ability to ‘see’ space in an intuitive and understandable manner for the ‘average’ user and lack in the interactive and media capabilities necessary to achieve an appropriate knowledge formation of the geography depicted. New methods and tools for the design of Virtual Environments have the potential to fundamentally change the role of maps in science and society (MacEachren et al. 1999). MacEachren has stated that Virtual Reality (VR) technology produces a more complete perceptual experience than traditional (carto)graphic (sic.) representations. VR offers a range of benefits for portraying geographical information. Three dimensions can provide a means for intuitive organisation of spatial objects that aims to replicate or reflect an abstraction of the real world. This virtual representation in turn utilises the user’s natural perception and memory of space and spatial relationships (Boyd Davis et al. 1996). The interactivity and dynamics of VR can stimulate the user’s engagement and understanding of the real world. By this reasoning, an appropriate modeling environment could come from a popular form of VR, a game, or the underlying driver for games, a gaming engine.
The approach for this research is based on the “brief” that advances in games and digital entertainment has been swifter than multimedia tools. Also, games are a familiar medium for users, with an estimated 75 percent of people under the age of thirty having played a computer game (Bryce & Rutter 2001). Nack (2001) argued that this situation was not chiefly responsible for establishing a competitive economic environment, where products thrive with successes and die with flops. Rather, it seems that the underlying question behind the development of games is somewhat different. In digital entertainment, the customer dictates to developers and sales personnel and developers constantly assess who their customers really are and whether their products meet customers’ needs. Multimedia researchers, however, tend to focus on how to solve a particular problem and therefore the real-world applicability element of the equation may be completely ignored. This is the fundamental reason why the skills of ‘non-expert’ users will be identified as part of this research program – to produce a GeoVE that utilises the full technological extent of contemporary digital modelling environments. This will lead to the “non-expert” user gaining a better understanding of a real place being depicted.
This research focuses on both geospatial technologists and the wider general public alike, offering an accessible and understandable tool with which to visualize landscapes. For this to be realised, it is intended that the visualization system will be offered using a desktop PC (as opposed to fully or semi-immersive systems) and to further engage user interest the virtual world will be built using games technology. The development of virtual environments as learning tools, based on games technology and applying a gaming metaphor could enable the development of a new pedagogical environment where participants can explore virtual landscapes and geographical knowledgescapes with a newfound confidence in wayfinding and navigation.
NAVIGATION AND WAYFINDING IN RELATION TO VIRTUAL ENVIRONMENTS
It has been reported that many users of desktop GeoVEs mention that they feel “lost” when using the product. It is also not uncommon for real world travellers to lose their way, especially in a city environment. The differences between becoming lost in the real and the virtual world differ somewhat though. For example, Lynch (1960) reported that most participants had trouble remembering components of the city in order to find their destination, therefore being “lost” is an issue of wayfinding. In virtual worlds the core problems associated with making users “lost” are linked to the Graphical User Interface (GUI) used for navigation, how it is used for orientation, the display space itself and how users related that display space to the geographic space it depicted (Slocum et al. 2001).
This may sound confusing, and this is due to the fact that often wayfinding and navigation are terms that are used interchangeably for everyday use. However, they are not the same when related to the use of virtual environments. Peponis et al. (1990) best described wayfinding as “the ability to find a way (from a starting point) to a particular location in an expedient manner and recognise the destination when reached”. Whilst navigation is most often defined as “the process of determining a path to be travelled by any object through the environment” (Darken & Sibert 1993). Orientation is our awareness of the space around us, including the location of objects and places. Thus it facilitates the understanding of the relations between current and target location (Hunt & Waller 1999; Downs & Stea 1977). Elvins (1997) concluded that “without wayfinding a navigator won’t know in which direction to steer and without navigating, a wayfinder will not have the means to move toward his destination”.
Over recent years it has often been the belief that fluid navigation and orientation relies on defining the appropriate metaphoric approach. The main role of metaphors in GUI design is to afford ways of interacting with the, often abstract, computer environment and to help users master complex tasks (Kuhn 1995; Shneidermann 1992). It is long believed that the map or earth metaphor is all that the user requires to make navigation easy in GeoVEs. Fuhrmann et al. (2001) disagreed with this, having assessed an improved “flying saucer” metaphor. However, their results also indicated shortcomings, “providing both advantages and disadvantages” having described that “user-domain-specific designed interaction metaphors support virtual navigation in desktop geovirtual environments but do not necessarily facilitate orientation”. Interesting, they refer to gaming based solutions, “participants suggested that the arrow keys
and other letter keys on the keyboard could be used for navigation”. This comes about when navigating through most GeoVEs via mouse movement. Also, “Many participants had experience with videogames and this experience may explain the quicker learning curve. The differences in our second and third focus groups support the contention that it will be important to investigate implications of the “Nintendo Generation” (Cartwright 1999) for future design of GUIs for desktop geovirtual environments.” Indeed, the gaming metaphoric approach offers highly promising navigation and orientation methods in comparison to all previous GeoVE navigation metaphors. Its advantages are amplified due to the widespread popularity of PC computer games, whereby millions of users have already inadvertently become navigation savvy using the game metaphor. Also, it is interesting to note that some VRML browsers now offer game mode movement facilities (i.e. navigation through the use of arrow keys or other designated keys and orientation can be made by mouse movement (Bitmanagment 2005).
THE PROTOTYPE ENVIRONMENT
The ‘proof-of-concept’ prototype will model the town of Queenscliff, Victoria, Australia (Figure 1). This historical township covers approximately 1.5 square kilometres and it is located on the eastern tip of the Bellarine Peninsula. It is a coastal town boasting historic buildings, situated on top of a slightly undulating landscape. Queenscliff was chosen because of its scenic backdrop, aesthetics and its aerial coverage is a reasonable modelling size for a gaming engine to handle. Queenscliff contains an historic past that the town has preserved, which possibly eliminates some problems associated with temporal mapping and visualization. Also, Queenscliff is the location of previous geovisualization research initiatives undertaken by the School of Mathematical and Geospatial Sciences, RMIT University, which has simplified the gathering of data required for this research.
Figure 1: Aerial photo of Queenscliff, Victoria.
Careful deliberation over the appropriate 3D gaming engine was paramount. A comparison was made between several high-end rendering games engines in order to select the appropriate visualization tool. The factors used in selecting the gaming engine primarily took into account how best to overcome the issues related to user discontent with current GeoVEs - ease of use, wayfinding and navigational functionality and graphical and interactive attributes. The authors also required certain attributes additional to user needs, principally choosing software that allowed for customisation of the package for creating representations of reality and actual cost of the development software, as distinct from user software costs.
The best-known subset of game engines is 3D First Person Shooter (FPS) game engines. Most groundbreaking development, in terms of visual quality, has been done on FPS games. Even today, despite the increasing realism of flight and driving simulators and real-time strategy games, First Person Shooters are still at the forefront of computer graphics. Currently there are four distinct generations of 3D technology employed in FPS games. These extend from planar worlds demonstrated by rectangular grids in Wolfenstein 3D, sector-based plane levels in Doom with sprite objects, through to games seamlessly integrating indoor-outdoor environments, pixel shader-based textures, bumpmapping, vertex shaders used for animations, lighting and shadowing. It was decided that in-depth
analysis of third and fourth generation FPS games would allow for the most appropriate working environment to be identified.
An analysis was conducted on several different types of FPS engines - Epic’s Unreal Runtime 2, id Software’s Quake III, 3D GameStudio, Criterion’s RenderWare, String Collaborative Virtual Environment (CVE) (an altered version of GarageGames’ Torque Game Engine) and Crytek’s CryEngine Sandbox. After a feasibility analysis was conducted three main choices that provided an appropriate working environment were identified. Those options were Unreal Runtime 2, String CVE and CryEngine Sandbox. After an initial implementation of the Queenscliff elevation model and the development of several static objects, it became apparent that the most stable, easiest to learn and most powerful engine was Crytek’s CryEngine Sandbox. The popular game developed on this engine was Far Cry, one of the first fourth generation FPS, launched in 2004.
When modelling a virtual representation of a real world environment such as Queenscliff, the game engine must be able to: Model a true 3D environment; Model landscape as well as architecture; Be accessible over a desktop PC; Allow the user real-time movement around the virtual environment; Allow the user to interact with the model; Provide animation and spatialised sound; and, Offer powerful graphics quality without diminishing system performance to an unsatisfactory level.
CryEngine Sandbox is a real-time multiplayer environment and editing application written entirely in modular C++. This meticulously designed fourth generation game engine pushes the threshold of action gaming with proprietary ‘Polybump mapping’, advanced environmental physics, Lua based event scripting, autonomous AI system, dynamic lighting and shading, motion-captured animation and total surround sound. The terrain uses an advanced height-map system and polygon reduction to create massive, realistic environments. The view distance can be maximised to an unprecedented 2 kilometres when converted from game units (Ubisoft 2005). (This exceeds greatly the view distance available on other products). Geometry, surface, and sound are authored in external applications and accessible via CryEngine to compose a map. This involves working with a range of software, and a typical workflow for the software employed is outlined in Figure 2.
The typical workflow begins with the collection of the base data, which is used to formulate the modelling environments. The base data consists of a Digital Elevation Model (DEM), imagery of textures, sounds, buildings and three-dimensional objects, avatars and animated objects. All of the aforementioned base data needs to go through a process that transforms it into an appropriate form that the game environment understands. Textures need to be the correct size (base 2 up to 1024 by 1024 pixels) and tile seamlessly before being packaged in what is known as a .dds file. Sound may further be manipulated in a package such as Audacity. ArcGIS allows the DEM of the town of Queenscliff to be visualised in three-dimensions and an image map can be generated, in preparation for entry into String. Buildings, trees, characters, animated objects and other three-dimensional entities are modelled using 3ds Studio Max.
After the base data has been processed, it can be imported into the Far Cry visualization environment. This is where the user will ultimately visit and explore Virtual Queenscliff. Alternatively, the user may view it via CD-ROM or possibly access it via an online multi-person environment. The appearance of Virtual Queenscliff in the Far Cry environment is shown in Figures 3 and 4. It represents a photo-realistic, true, three-dimensional environment, allowing the user to control their navigation within the space. This space may include perspectives not achievable by reality, such as subterranean areas or inaccessible aerial views. Given that this environment allows for real-time movement and interaction, the CryEngine Sandbox engine offers a power and ease-of-use that geospatial technologists and cartographers are beginning to widely accept (Moloney 2003).
DDS plugin
NVIDIA or other .dds .JPG & .PNG compiler: ref. images to Photoshop or equivalent enable use in 3ds Max graphics software: Graphics generation for use in CryEngine, 3ds Max.
(.dds or .bmp files) 3ds Studio Max: Constructive Solid Geometry & character and shape modelling / animation (.max files)
CryEngine Sandbox: Fractal based terrains and 3D ‘paint’. Animated water, atmosphere & particle systems.
A udacity or equivalent sound software ArcGIS or ArcView: (.wav files) DEM data input and heightfield map generation for terrain (.bmp files) Figure 2: CryEngine Sandbox workflow
A couple of examples of the final “look” of Virtual Queenscliff are shown in Figures 3 and 4 below.
Figure 3: Fort Queenscliff
Figure 4: Queenscliff Harbour
TESTING
The underlying theory for testing is the basic elements of the city, introduced by Lynch (1960). It does lack a formalized definition exact enough to be applied in a computational environment (Conroy Dalton & Bafna 2003). It can also be argued that the amount and conceptualization of the basic elements was not subject to a thorough and exact process, but rather an empiric identification. However, later works showed that the elements were identified correctly and distinguished the most important morphemes of urban environments. Lynch’s specification of basic elements of the city: paths, nodes, districts, edges and landmarks, are now generally accepted (Lynch 1960).
Taking advantage of the naturalness of a space-to-space mapping between real world and display does not require that GeoVEs replicate reality in all respects (Eiteljorg 1998). Realism can be a distraction. An air photo, for example, represents visual aspects of the world realistically, but does not function as well as a more abstract map for navigation. Similarly, an ultra-realistic virtual environment (by itself) may not function well as a tool to explore geospatial information. Thus, the same rules of abstraction and generalization relied upon for successful cartographic representation in two dimensions may apply in three. ‘Tapping’ into the iconic urban spatial cues is the philosophy behind Lynch’s theory for improved wayfinding and navigation.
Many researchers in the field of urbanism, architecture, anthropology and geography have tried to provide tools and methods to create models of city structure. This research uses games development tools to do the same. By developing Lynch’s five elements of design and applying his ‘environmental mapping technique’, this research evaluates the potential of using GeoVEs to improve ‘place legibility’ of a chosen urban environment. This will be done by initially defining areas of Queenscliff that are difficult to visualize through the ‘environmental mapping technique’ and by building a GeoVE that manipulates the geometry of the elements in the environment as well as their location in the virtual world. The next step will be to incorporate the five elements of design as augmented Virtual Reality tools for defining urban form, navigating Virtual Environments and improving spatial cognition of built environments. The final stage of the research will involve usability testing of the GeoVE urban design prototype, where ‘foreign’ participants to Queenscliff (i.e. test candidates who do not reside in the township) will comprise the test group, performing wayfinding tasks without Lynch’s elemental manipulations (Figure 5).
Locals draw image map of Queenscliff
Compare to original Test locals in wayfinding Combine and image maps and navigation tasks compare image through real Queenscliff maps – identify low imageable areas
Make locals redraw image map
Build altered VE of Build current Queenscliff to correct Queenscliff unimageable areas – use Lynch’s theory
Test on ‘foreigners’ to Test on locals as take out effect of measure of comparison Queenscliff between VE and real knowledge world
Compare results with Will results show real world navigation: improved imageability answers navigation / wayfinding? issues
Figure 5: Testing methodology
The users will be tested on how they complete specific wayfinding tasks. They may have to find a particular place and collect an artefact, or perhaps find a specific cannon in Fort Queenscliff and fire it to complete the task (Figure 6). Besides the implementation of Lynch’s elements, numerous other interactive events will take place that aid the user to complete wayfinding tests. An unaltered version of Queenscliff (without Lynch’s element manipulations) will be tested by a group unfamiliar with the town of Queenscliff.
Figure 6: Users may be asked to find and fire this cannon to complete a wayfinding task
The possibility of other modes of movement apart from walking is feasible. Users can drive vehicles (Figure 7) including land and sea vehicles, explore in fly mode (Figure 4), or even use experimental vehicular travel like hang gliders. Collision is used to constrain visitors’ paths through parts of the environment. Trigger events can set off voice descriptions of where you are and where you need to go. A flashing arrow located on the compass can also guide visitors to their destination or mouse-over functionality provides users with information when they want it, rather than having to follow a predetermined sequence. Positional sound can be used as navigational events and other acoustic and visual events randomly occur, or are triggers by certain user actions.
Figure 7: Explore Queenscliff by driving around it!
CONCLUSIONS AND FUTURE WORK
The advent of geovisualization technology has seen the development of a number of tools that can support urban planning processes including: Computer Aided Design (CAD), Geographic Information Systems (GIS), multimedia/hypermedia products and now three-dimensional gaming engines. With computer gaming tools now readily available for deployment from desktop computers there now exists the opportunity to possibly extend traditional spatial cognitive theories and test their applicability using Geospatial Virtual Environments (GeoVE).
In the current research agenda of the ICA commission on visualization and virtual environments, the development of methods to assist users in navigating and maintaining orientation in GeoVEs has been identified as one of several major research challenges (MacEachren & Kraak 2001). This research introduces the idea of gaming environments and the use of the gaming metaphor as an improved form of navigating and orientating oneself in a GeoVE, one which effects parts of the environment along the lines of Lynch’s elements in order to improve ‘place legibility’ and thus improve wayfinding. Although not subject to a thorough or exact process, Lynch’s theories have proven to be correct over the years, and recent research by the likes of Strohecker (1998), Klatzky et al. (1997), Ingram and Benford (1995) and Dieberger (1994) has revisited Lynch’s theory, applying Virtual Environment technologies to widen its usability. The basics of the research outlined in this paper are an extension in a comprehensive investigation in navigation, orientation and wayfinding research, directed to facilitating travel in desktop GeoVEs. Further work is required to evaluate the effectiveness of the gaming engine as a GeoVE tool for better understanding the image of a city, applying the testing procedure discussed.
ACKNOWLEDGEMENTS
The authors wish to acknowledge Crytek for providing use of CryEngine Sandbox, the accompanying Software Development Kit, and the online forums “CryMod” and the “Official Far Cry Website Forum”. The researchers would also acknowledge RMIT University and the financial support provided by an Australia Postgraduate Award (APA).
REFERENCES
Bitmanagment 2005, Bitmanagement Software Developer Information, Bitmanagement Software GmbH, 29 April 2005,
Lynch, K. 1960, The Image of the City. MIT Press, Massachusetts. MacEachren, A. M., Edsall, R., Haug, D., Baxter, R., Otto, G., Masters, R., Fuhrmann, S. & Qian, L. 1999, 'Exploring the Potential of Virtual Environments for Geographic Visualization', in Annual Meeting of the Association of American Geographers, Honolulu, Hawaii, March 23-27, 1999, AAG, pp. 371-380. MacEachren, A. M. & Kraak, M.-J. 2001, 'Research Challenges in Geovisualization', Cartography and Geographic Information Science, Vol. 28, No. 1, pp. 3-12. Moloney, J. 2003, String CVE, June 12,