Generative Systems for Design, S22

This course is for visual artists, enthusiasts, and designers from different areas (game, product, architecture, building performance, etc.). Taking a stance, say of, instead of working directly with analog media, such as drawings, designers can develop algorithms and computational models that can generate design alternatives, based on custom input. We refer to such algorithms and models as generative systems.

Generative systems have been an important topic of research in recent decades, in books, courses, articles, and research conferences in computational design and in other disciplines. Earlier approaches were based on classical artificial intelligence and optimization methods; recently, a variety of computational techniques from different fields, such as parametric modeling, agent-based modeling, or neural networks, have been incorporated in the development of new generative systems. With recent developments in machine learning, we can even develop models that learn automatically from data or experience. In this course, we focus on the basic or classical techniques, although with an acknowledgment to recent developments, to briefly introduce learning techniques.

The main goal of this course is to foster the student's capacity to formulate design problems computationally, with emphasis on the synthesis of design alternatives. This course provides an overview of the main topics in Generative Systems, with historical notes and technical specifications. Throughout the semester, the students will address different design problems with different generative techniques. The course will address topics such as variational modeling, rule-based modeling, enumerative composition, cellular models, agent-based modeling, and optimization. The appropriate data structures, algorithms, and models will be discussed, and some implemented in the exercises and projects.

Adversarial System for Building Generation, Sachin Dabas, 2022

The course has six different modules by different generative approaches in design. Module 1. Variational Modeling | Module 2. Rule-based Modeling | Module3. Enumerative Composition | Module 4. Cellular Models & Agent | Module 5. Agent-based Modeling | Module 6. Optimization 

Maze City Generation, Chloe Wang & Harley Guo, 2022

Each module is consisted of 2-3 lectures and 1-2 recitations for 2 weeks. Lecture will cover each approach’s basic concepts, motivation, usage, algorithms, and data structures with a short lab session which demonstrates how the generative algorithms works in design problems. Expanding the lab sessions of the modules, recitations will provide opportunities to understand the algorithms in detail.

Mass Builder for Urban Point Clouds, Bradley Williams, 2022