Terrain Database Requirements Become More Demanding
Written by Erin Flynn Jay
MT2 2010 Volume: 15 Issue: 2 (April)
Nearly all aircrew trainers include some type of image generation and display system, which may provide simulated out-the-window scenery, video camera systems, radar, night vision displays, infrared sensors, etc., said Colonel John Franz, commander, 677 Aeronautical Systems Group for the U.S. Air Force. These systems are tailored to enable the conduct of specific training tasks assigned to the simulator, such as takeoff/landing, navigation, weapons employment and aerial refueling. In order to provide effective training, it is often critical that the displays presented to the aircrew member are realistic and provide an accurate depiction of the operational environment he or she would encounter in combat, Franz said.
The key to achieving realism in a simulator display is the creation of a believable representation of the real world, whether it is the aircrew’s home training base or a combat theater. This is accomplished through the construction of a terrain database, which consists of a mathematical model of the earth’s surface, as well as the natural and man-made features on it. Franz said the specific content of this database is defined by the training tasks to be performed.
If an effective training scenario relies on the accurate representation of certain elements that are unique to a particular geographic location, a geospecific database may be required. “A common requirement which drives a geospecific database is the need to train takeoffs and landings at the aircrew’s home base. For this application, the base’s runways, taxiways, control tower, lighting system and other airfield structures must be modeled exactly as they are in the real world,” Franz told MT2. “The planned use of a training device to conduct mission rehearsal is another major reason for requiring a geospecific database.”
Geotypical data can be used in many circumstances when real-world truth data is not necessary, but there is a desire to include scene detail that is generally representative of an actual geographic area. This may be a suitable solution in desert scenarios, for example, where no geographically significant features exist, i.e., “all deserts looks alike.” Since less source material must be procured, Franz said populating a database with geotypical data is often much less expensive than using geospecific data, so it is used whenever it is practical to do so.
As image generation technology becomes more sophisticated and computer performance increases, the requirement for terrain databases becomes more demanding. Terrain, as well as models of 3-D objects such as vehicles and buildings, is rendered as polygons; as computer processor performance increases, it becomes possible to render increased numbers of polygons in each video frame. Operationally, Franz said this translates into the trainee being able to see and interact with a more highly detailed environment, thus making the simulator training more effective.
From a database perspective, the enhanced capability drives the need to develop more comprehensive visual models, requiring additional polygons to provide this detail. The greater information content creates a corresponding increase in the cost of building databases. “Likewise, higher pixel resolution displays enable aircrew members to view increasingly finer details, which drives the need for higher resolution imagery to provide texture,” said Franz. “This need for higher resolution source material is directly analogous to the technology advances which occurred in the home television industry— when the data content of existing DVD discs proved inadequate to support HDTV, the higher resolution Blu-Ray disc format was developed.”
DATA SOURCE CHALLENGE
The lack of a common data source can be a challenge, Franz said. The National Geospatial Intelligence Agency (NGA) is the largest source of data for DoD, providing the majority of the terrain elevation, vector feature data and imagery; however, many trainer programs also acquire imagery products from commercial sources. Many of these sources are standardized, widely available, and straightforward to process using automatic methods.
Conversely, nonstandard sources such as drawings and photographs provide the source information required to construct 3-D models. This method requires a significant amount of manual interaction to create an acceptable representation of a real-world item. “This makes it very time consuming and expensive to create complex visual models, such as aircraft, and the resultant models cannot be easily shared among simulators,” Franz said. “Another challenge is the ability to obtain adequate source material for highfidelity simulations which do not operate in the visual region of the electromagnetic spectrum, such as radar and infrared sensors.” Because geographic databases represent a significant portion of the cost of developing and maintaining simulators, all of the department’s services have recognized the need to standardize geographic information among systems. The Air Force Common Dataset is an example of one of these initiatives. By requiring new trainers to accept and exchange geographic information in a limited number of widely accepted standard formats, it becomes possible to share data among systems, thereby reducing the overall cost of database acquisition and support to the government. The 677 AESG has begun requiring Common Dataset standard compliance on its most recently awarded contracts, and will eventually be able to benefit from the reuse of data across platforms.
MetaVR Inc. has not built terrain based on geotypical imagery for several years now. “As far as we can tell there is no market demand from our customers for terrain built with geotypical imagery,” said W. Garth Smith, president of MetaVR. “Geospecific imagery is so readily available and our workflow is such that would be far more work to build geotypical terrain than it is to build geospecific terrain. Our tools are automated for terrain and imagery construction.”
Smith said the sources of MetaVR’s geospatial data for its products are Kirtland Air Force Base Simulator Database Facility, Digital Globe and Aerial Express.
MetaVR is working on product improvements in higher resolution imagery and elevation data, and higher quality and density of geospecific content. “Imagery and elevation is upstream of our workflow, and we are typically limited by the quality of source data. However source imagery and elevation data continually improve and decrease in cost,” said Smith.
MetaVR’s workflow for terrain content is based upon 3-D Studio Max and has undergone significant improvements, particularly with the development of their Afghanistan terrain. “Our plan is for our customers to see more 15 cm imagery and DTED [digital terrain elevation data]3 (10 meter terrain posts), as well as areas on the order of five times the size of our current Afghanistan terrain over the next two years,” Smith concluded. “Customers will see more game-like content on buildings and street detail, as well as modeled building interiors. We are also integrating deformable buildings and terrain into our next generation terrain.”
DIFFERENT REPRESENTATIONS
Dave McKeown, president of TerraSim Inc., said that visual systems and constructive simulations have fundamentally different representations for the same geospatial area. While visual simulations are concerned with efficient display of multiple levels of imagery, terrain and 3-D models, constructive simulations require a much larger amount of semantic information or feature attribution. Attribution is directly associated with the geometry and appearance of surface material (mobility) and both natural and man-made objects. Constructive simulation runtimes rely on accurate attribution to perform planning and to reason about the 3-D environment.
A significant challenge is correlation between visual and constructive simulations, McKeown said. Early systems could use the same polygonalization to represent features across both representations. Basically, one representation could be translated polygon by polygon into the other. A good example would be the relationship between a CTDB [compact terrain database]-based constructive simulation (OTBSAF, JointSAF) and a visual system, which would use the same terrain skin, but apply a phototexture to the visual surface and place models on or integrated into the surface.
However, as visual systems using serious game technology have come to be more prevalent, there are fundamental differences between the underlying representations that make a translation approach incompatible with the goal of correlation. “Many of our customers are interested in using OneSAF and VBS2 in networked simulation. OneSAF used an integrated TIN [triangulated irregular network] approach, where each surface triangle is required to store attribution as to surface material, mobility and other properties used by the OneSAF runtime,” McKeown said. “The OneSAF OTF format is defined as a geodetic TIN with elevation referenced to the earth’s ellipsoid. VBS2 and most gamebased simulations use a regular grid spacing, typically using UTM [universal transverse mercator] or a local rectangular coordinate system, with elevations referenced to the earth’s geoid.”
TerraTools produces the OneSAF OTF surface and content and then generates the equivalent VBS2 surface and content model, taking into account the inherent differences in the two runtime representations.
AFRL INTEGRATES TWO SIMULATORS
An innovative approach to striking the balance between high fidelity and operational integrity is being demonstrated by the Air Force Research Lab. AFRL’s Warfighter Readiness Research Division of the 711th Human Performance Wing, Human Effectiveness Directorate (711 HPW/ RHA) undertook a novel seedling effort where it integrated or combined two existing, freestanding training simulators, making them interoperable with each other for live, virtual and constructive training. The goal was to demonstrate the rapid integration and application of existing and relevant technologies for a new training requirement without creating a new training system out of whole cloth.
The two simulators AFRL identified for this integration effort were Aptima’s DDD (Distributed Dynamic Decisionmaking synthetic task environment), a desktopbased training simulator that allows users to quickly and easily design training scenarios, and L-3 Communications’ Link Simulation & Training’s Blue Box HD Training System, a simulation environment that provides immersion in a complex high definition living world.
AFRL demonstrated the DDD-Blue Box integration at the 2009 I/ITSEC. It featured a DDD-created humanitarian disaster response scenario populated with entities representing emergency responders, victims, firetrucks and other rescue resources. This rescue and support scenario was displayed within the Blue Box high definition virtual world, which creates geospecific virtual simulations using dynamic visual system database content.
“The combined simulation showed the DDD-scripted scenario and entities interacting within the Blue Box high definition computer generated imagery. Integrated within the scenario was a remotely piloted aircraft (functioning as a UAV), which provided the trainee an overhead “eyes in the air” perspective for an exceptional level of training realism and operational situational awareness from which to rehearse and manage the disaster relief mission,” said Michael J. Paley, executive vice president, Aptima Inc.
The DDD-Blue Box integration addressed the balance between the ability to create mission-relevant scenarios with an appropriate level of fidelity for training usefulness/ effectiveness. While high-fidelity training environments provide realism, they often do so at the expense of being too complex and inflexible to modify for the desired scenario. Paley said DDD’s strength is that it can be easily configured to simulate a wide range of team-based scenarios and environments, such as AWACS air battle operations, joint task force command, and emergency disaster management.
Both virtual and constructive simulations experience similar challenges, but virtual simulations are most demanding since they must rapidly produce environmentally realistic scenes with near instantaneous response to human and artificial inputs, said Craig Fones, M&S account manager at ESRI Inc. Image generators need specially prepared run-time databases in order to meet demanding dynamic 3-D display performance, highly detailed graphics and environmental effects, and must do so while synchronized across a network of gaming systems supporting various levels of detail.
“What they have in common is that they start out with terrain data sources that are common across the intelligence and operational disciplines, but then must restructure that data to meet the particular needs of a given simulation,” said Fones. “This altering of the source data can induce many data problems related to correlation issues, such as dissemination, management and upkeep, to satisfy the demanding display requirements of image generators.”
A recent trend is to merge geographic information systems (GIS) with simulation applications to provide solutions that utilize GIS source data directly, requiring little or no data processing, or GIS software as a core component in constructive simulations. Virtual simulators are more demanding of data sets, but today’s graphics processors are making high-performance 3-D rendering directly from source data and GIS software a reality. Fones said that close integration between GIS and modeling and simulation (M&S) software is a quickly evolving technology that enables defense organizations to train using source data and geographically correct terrain.
There will be a major release of Arc- GIS 10 this summer, offering better tools to manage imagery in the enterprise and exploit imagery in a GIS workflow, intuitive 2-D/3-D vector editing and data production tools, concluded Fones. Data management will be simplified and best practice templates will be readily available to provide the user with proven examples to jump-start projects. ♦