Beschreibung

The modular system can be used in modern building construction in order to achieve a great architectural variety with as few module variants as possible (e.g. columns and beams). To achieve this goal, optimization algorithms are used that are based on the geometric data of existing buildings. As the construction kit is designed for the widest possible building variants, the building data must be extended by artificial data.

Generative Adversarial Networks (GAN) offer one way of generating this data automatically. For the application of a GAN, the two underlying networks, the generator and the discriminator, must be adapted to the problem. The building data are represented by graphs, which are initially generated on the basis of random manipulation of parameters. The building data generated in this way forms the data set for training the GAN. 

Aufgabe:

• Generation of a suitable training data set
• Literature research on generative adversarial networks in the context of graphs
• Implementation, comparison and evaluation of different approaches

Profil:

• Ability to work independently
• Interest in product development, optimization and 3D modeling
• Basic knowledge of programming

If you are interested or have any questions, please contact me: niklas.frank∂kit.edu