October 2022 – May 2023
Research Partners:
- Department of Computer Science, Faculty of Mathematics and Computer Science, University of Bucharest (CS-FMI-UB)
- Ion Mincu University of Architecture and Urban Planning (UAUIM)
- International Center for Research and Education in Innovative Creative Technologies (CINETic)
- “Arhitectura Celuilalt” Association
Research Team:
- Andreea Robu-Movilă (UAUIM) (Principal Investigator)
- Sergiu Nisioi (CS-FMI-UB) (Principal Investigator)
- Dorin Ștefan (UAUIM)
- Sabin Țenea (UAUIM)
ILDS Management Team:
Keywords:
Evolutionary-Genetic Algorithms, Form Finding, Design Optimization, Heuristics, Metaheuristics, Human-Machine Collaboration, AI-driven empathic computing
Description:
The proposed hybrid optimization model for architectural design processes is based on coupling evolutionary-genetic algorithms (Biomorpher and/or Wallacei) — to address form generation; metaheuristic algorithms (Cluster-Orientated Genetic Algorithm – COGA, and NSGA-2) — for classification according to the performance criteria; and “affective computing” (Picard 1995) — for human curation and feedback. This model enacts the fluency of the design process through “direct manipulation” and promotes a co-creative interaction between man and computer that minimizes what Hutchins (1985) called the “gulf of execution” and “gulf of evaluation”, namely the latency between ideation, representation, and evaluation. When the form evolves autonomously, the architect is no longer the absolute depository of the creative act, therefore a fluid collaboration through affective computing would constitute the premises of what Makato Sei Watanabe called the “algorithm of preferences”, thus a “Grey Boxing” (Andrew Witt), a controllable variation of the mechanical “Black Box”.
Methodology:
In the first stage of the experiment, we will analyze the complex decision-making processes inside the designer’s mind by scanning the brain’s emotional feedback described by two parameters, arousal & valence. These indicators can be measured using neuroscientific methods and our framework is based on a EEG (electroencephalography) headset Emotiv EpocX composed with eye-tracking response to outline which solutions the eye fixates longer onto. This signals will be interpreted by NeuroPype platform that will identify the affective activation pattern associated with the mechanism of decision-making. Based on the design proposals that unconsciously show more interest/motivation for the designer, a feedback loop will be activated to train back the algorithm for generating new iterations.
The second part of the experiment will run an algorithm which will only be metaheuristically optimized and the obtained results will be analyzed comparatively in terms of process and outcome.