Research Activities

Standardization, Digital Building Permit, and AI (NeRF and Generative AI for geometry)

Standardization

At RDF Ltd. we are implementing state-of-art support for several open standards related to geometry: AP203, AP214, AP242 (till v3), CIS/2, IFC (2, 2x ADD1, 2×2 ADD1, 2×3 TC1, 4 ADD2 TC1, 4×1, 4×2, 4.3 ADD2, 4.4 and 5), IGES, Collada, glTF, LandXML, CityGML, OBJ, GEOM ontology and several others. Because of this experience and to understand these technical standards in depth, we participate in the definition of several of these standards. From inception till official CEN / ISO certification.

Some of our successful research work concerns the co-development of the technical part of IFC4.3 ADD2 with clear alignment definitions in both domain-related representation as well as the geometrical representation; with a clear definition of how to sync redundant parts, that can both exist independently. Based on the Semantic Web we developed in the past two decades the GEOM ontology containing geometrical primitives that can extended parametrically with user-defined primitives in a distributed manner fully ‘understood’ by our Parametric Geometry Modelling Kernel.

Digital Building Permit

There are many interesting and relevant research areas for the architecture, engineering, construction, and operation (AECO) domain. On the border of our expertise on semantically rich geometrical representations and deriving knowledge from existing non-geometrical semantics there is the area of rule definition and automated rule checking. Next to the challenge of defining such rule (related) knowledge in an open and reusable manner (for example through SPARQL / SHACL), there is often the situation that knowledge is available in different (open) formats. For example IFC for buildings and CityGML for surroundings. RDF Ltd. is involved in the CHEK project and is closely following the developments in the projects ACCORD and FireBIM.

AI (NeRF and Generative AI for geometry)

As nearly any self-respecting software company we closely follow AI developments and texting promising LLM’s. Practical functionality is used, although (for now) we do not use AI for populating any code within our commercial libraries.

Where we try to be one step ahead of the market is in relevant AI developments in the areas of our main expertise, i.e. semantically rich geometry definitions. At the moment this relates to using NeRF (Neural Radience Fields) to generate fully automatic 3D textured meshes from 2D content. Using Semantic Web inferencing techniques in combination with AI to automatically split meshes, enrich the semantical representation of this mesh from the technical side as well as from the classification/language side. Also, we use Generative AI on top of existing LLM’s to enable the automated generation of proper geometrical design trees based on our primitives in the GEOM ontology and newly created primitives through the parametric extension mechanism from our Parametric Geometry Modelling Kernel.