Back to ASAM International Conference 2019

 

 

Development of a 3D model exchange format with physical material properties for virtual development, test and validation 

 

Dr. Ludwig Friedmann 
Solution Architect Simulation Autonomous Driving
BMW AG 

 

In virtual development, test and Validation of automated and autonomous driving functions, 3D models are used for the geometric representation of the environment of simulated vehicles as well as for the vehicles themselves. For a long time, physical correctness in the visual representation of those geometries was not fundamental. Due to limited computing capacities, implementations were designed for lowest possible memory and computing time requirements while providing a visually plausible appearance. 


For physical sensor Simulation, which is becoming increasingly important in the context of virtual develop-ment, test and Validation of automated driving functions, this approach is suitable to a limited extent. Instead of visual plausibility, physically correct modelling of material properties is fundamental in order to achieve valid results. Besides sensor Simulation, modern rendering solutions also require physical material properties in 3D models in order to be able to reproduce physically correct lighting, reflections and shadowing. 


The specific architecture of virtual development, test and Validation frameworks poses further demand for action. Within the frameworks, which may be set up distributed over several compute nodes, subsystems such as the environment Simulation, rendering and sensor models are implemented as individual software components. Internally, these components use non-standardized 3D models to represent the environment and road users, each for its specific purpose. Examples are visual representation as well as the calculation of wheel contact points and the propagation of beams or wave fronts. The implementation can be both in the form of a white box or a black box and may originate from different suppliers. 


A 3D model exchange format, physical material properties and their annotation in corresponding 3D models are not yet standardized in this context. This results in ongoing integration efforts as well as in incompatibilities among software components. Yet, it is not possible to guarantee the consistency of 3D models in environmental Simulation, rendering and sensor Simulation. Moreover, the lack of uniform material definitions and annotations prevents the creation of comprehensive 3D model databases and material databases for virtual development, test and Validation. For the comprehensive use of physical sensor models, which requires extensive, uniformly defined environmental geometries and associated materials, this standardization measure is indispensable. 


Within this project, a prototypic implementation of a Specification for the standardized annotation of physical material parameters in the open 3D model exchange format glTF (github.COM/KhronosGroup/glTF) is developed. glTF is developed at the Khronos Group (www.khronos.org/) and already has a broad following in computer graphics. 


The implementation incorporates an exemplary 3D model including annotated material properties for common automotive sensors and serves as Specification. A sensor model that queries the material information of the 3D model is implemented as an automated test. 


The project was launched with the aim of publication. The publication of the implementation should serve as a template and as basis for discussion. In the midterm, it is planned to transfer the results to standardization committees in order to create a uniform basis for virtual development, test and Validation and a foundation for the development of comprehensive 3D model databases and material databases.