KityVR - Artificial intelligence techniques to implement CityGML models and VR visualization

The development of 3D city models enables the integration of city and building levels into a single model that contains both semantic and geometric information. These models can be used to assist stakeholders such as policy makers, urban planners, utilities and energy suppliers, and citizens in making decisions, as well as to improve communication and outreach. The highest form of visualization in this field is the use of virtual reality (VR) technology. CityGML is an international, widely used Open Geospatial Consortium standard for modeling, storing, and sharing 3D city models; the Open Geospatial Consortium considers the fusion of virtual reality and CityGML to be one of the greatest challenges in 3D city model research and development. The KityVR project addressed two key research aspects in this context. (i) How can Virtual Reality be used as the highest form of visualization in the field of 3D city models? What are the use-cases for Virtual Reality in the domain of 3D city models? (ii) Modeling the energy structure of larger urban units is characterized by a lack of descriptive input data. In many cases, data sets are not complete or certain use cases require additional information that is not represented in the corresponding data model. In these cases, "enrichment" of the original dataset is necessary to provide all the required data. Machine learning and statistical methods are promising in this context: missing information is calculated using other attributes and heuristic approaches.

Results and conclusions

Results of an expert survey show that data acquisition is the most time consuming part of the research (median 40% of the total project duration). More than 80% of the experts use methods for data enrichment; approaches with building archetypes are used by 75%, statistical methods by 50% and machine learning methods by 36% of the respondents. Experts expect the greatest potential for machine learning in the area of data enrichment and analysis of input data. Three main areas regarding applications in the context of data enrichment could be identified: Parameter derivation or the replacement of missing data, the creation of more accurate building archetypes, and image analysis.

In the project, the simulation pipeline SHP2SIM was developed, which allows to directly create simulation models based on 2D building information.

In addition, a review paper analyzed different use cases of VR and their strengths and weaknesses. Furthermore, VR was implemented for two different applications in the field of buildings and cities; simulation results were visualized in VR.

Outlook

The fact that (i) data availability in the field of City Information Modelling is still limited, and (ii) more than 95% of the work in the field of City Information Modelling is not reproducible, shows that on the one hand, there is a need for clear political guidelines and a change in the legal situation, and on the other hand the scientific community has to critically question the quality of its own research.

Project management

Institute of Software Technology/Graz University of Technology

Project or cooperation partners

Graz University of Technology:

• Institute of Interactive Systems and Data Science
• Institut für Hochbau
• Institut für Wärmetechnik

Institute of Software Technology/Graz University of Technology
Gerald Schweiger
Inffeldgasse 16/b
A-8010 Graz
Tel.: +43 (316) 873 - 5747
E-Mail: gerald.schweiger@tugraz.at