Synthetic Environment Database Correlation

46 © 2015 Open Geospatial Consortium

1. Raw source correlation: Is the degree of informational consistency between

two or more sets of raw data 18 i.e., inputs to a modeling station representing aspects of the same environment for instance, the correlation errors arising from Digital Terrain Elevation Data DTED elevation data that does not perfectly match to satellite raster imagery due to oblique view distortions induced by the satellite. Correlation errors are intrinsic to the process of gathering data because since there is no means to gather all of the required data from a single device, at a single instant in time. Instead, datasets e.g., elevation, raster imagery, geometry are each gathered from various devices of various types with distinct precision, formats, capabilities, fidelity at different times. This in turn leads to a broad range of correlation errors typically resolved by the modeler during the final assembly of the synthetic environment from its sources.

2. Source database self-correlation: Is the degree of informational consistency

between the internal datasets of a source database produced by a DB generation toolset. To a large extent, the effort expended at DB generation time consists in eliminating or at least reducing correlation errors arising from miss-correlated raw source data.

3. Runtime database correlation: Is the degree of informational consistency

between two or more runtime client-specific databases representing the same synthetic environment 19 . The likelihood of achieving correlated runtime client-device databases is particularly low when different authoring tools and possibly different source data are used to assemble each of the compiled runtime databases. In recent years, some authoring tools have been improved to automatically produce a set of client-device database from one common repository internal to the tools. Nonetheless, it is still current practice within the simulation community to independently deploy the simulator client-device databases; as a result, correlation errors may occur especially if the master database repository is constantly evolving. The CDB Specification eliminates database correlation errors since only one database is used to represent the same synthetic environment. The CDB is a single database that can be accessed simultaneously by all simulator client-devices at runtime. By definition, it addresses all runtime database-level correlation errors.

4. Numerical accuracy correlation: Is the degree of informational consistency

between the outputs of two or more devices, with each device performing the same algorithms, using the same control parameters but performing internal computations to a different numerical accuracy. Consider for example two 18 In this context, raw source denotes any input to the modeling workstation that is used to assemble the synthetic environment; consequently, the data may have undergone some level of post-processing such as image color- balancing, image ortho-rectification, etc. or may be in a specialized source interchange format such as SIF, SEDRIS, etc.. 19 A runtime client-specific database is a device-loadable database format that can be processed by a target device. 47 © 2015 Open Geospatial Consortium devices computing the sin of an angle, one with a series of 10 terms, and another with an interpolation of a look-up table with 100 entries, or one device using 32-bit signed integers for its internal computations and the other using single-precision floats. The CDB Specification reduces numerical accuracy correlation errors because a single representation is used for each dataset. Note however that numerical correlation may deteriorate due to the variances in numerical precision inherent to each client-device.

5. Algorithmic correlation: Is the degree of informational consistency between

the outputs of two or more devices, with each device performing its internal computations to the same numerical accuracy, but using different algorithms with possibly different control parameters. Consider for example, two devices meshing terrain from a regular grid of elevation points, one using a regular mesh of right-handed triangles using the elevation points as vertices, and the other with a DeLauney triangulated mesh derived from the grid of elevation points. Algorithmic correlation errors can be introduced anywhere in the dataset processing chain, ranging from the raw source level right through to the internals of the simulator client-devices. While it is not possible to mandate complete algorithmic uniformity throughout this elaborate chain, the CDB Specification offers solutions to this issue 20 . Firstly, only one database is produced, so all DB authoring tool algorithmic correlation issues are eliminated. Note that the implementation of runtime publishers on a simulator can play a role in improving overall algorithmic correlation 21 .

6. Parametric correlation: Is the degree of informational consistency between

the outputs of two or more devices, with each device performing the same algorithms with identical numerical precision with different control parameters e.g., consider two devices generating regular meshes of right- handed triangles based on a regular grid of elevation points organized by LOD, one using an LOD meshing tolerance parameter of 1m and the other one using 2m. Note that it is clearly possible to create a synthetic environment database whose content is at a level of resolution, fidelity and accuracy that can overwhelm any client-device; however, the amount and type of data that is rejected by each type of client-device can vary considerably. For example, a NAVAIDs data server client-device need only be concerned with Digital Aeronautical Flight Information File DAFIF™, and as a result is virtually unaffected by changes in the resolution of the terrain elevation data which could easily increase terrain SE content by 100-fold. The implementation of the CDB Specification on a simulator can reduce andor eliminate parametric 20 For instance, the CDB does not impose the terrain skinning and meshing conventions that each of the runtime publisher generates for its attached client-device. However, it is understood that if all client-devices would use the CDB meshing conventions natively, correlation issues between clients would be significantly reduced. 21 Since each dataset is uniquely represented, it is possible to share a greater number of algorithms between CDB runtime publishers. This ensures that datasets are processed in an identical fashion whenever two or more publishers share the same algorithm.