ARA - Grasshopper Plugin for AI-Augmented Inverse Design


METADATA ONLY
Loading...

Date

2024

Publication Type

Conference Paper

ETH Bibliography

yes

Citations

Altmetric
METADATA ONLY

Data

Rights / License

Abstract

We introduce ARA - a Grasshopper plugin that integrates AI-augmented and data-driven design with parametric modelling, enabling architects, engineers and designers to efficiently generate design solutions with the assistance of generative neural networks. The inverse design paradigm accelerates design exploration by providing many different design variants that match requested objectives. In this paper, we present the main features of ARA and demonstrate its application in two design scenarios. The presented workflow allows to create project-specific solutions, which includes generation of the dataset, training and deployment of a custom autoencoder model to generate designs, and various visualisation tools for data analysis and performance evaluation. We also briefly explain the underlying machine learning methods and provide some commonly used best practices. The main motivation for this work is to lower the barrier to entry for users without machine-learning background and to accelerate adoption of deep-learning methods for everyday design tasks.

Publication status

published

Book title

Scalable Disruptors

Journal / series

Volume

Pages / Article No.

231 - 240

Publisher

Springer

Event

Design Modelling Symposium (DMS 2024)

Edition / version

Methods

Software

Geographic location

Date collected

Date created

Subject

Inverse design; Computational design; Generative AI; Machine learning; Autoencoders; Grasshopper for Rhino

Organisational unit

03709 - Kohler, Matthias / Kohler, Matthias check_circle
03708 - Gramazio, Fabio / Gramazio, Fabio check_circle

Notes

Funding

Related publications and datasets