IAAC – Master in Robotics and Advanced Construction
Workshop 3.1
Faculty: Zeynep Aksöz
Faculty Assistant: Ashkan Foroughi, Ricardo Mayor

LEARNING TO BE A SPACE FRAME


Credits:  Zeynep Aksoz

Syllabus

The involvement of machine intelligence enables a new design optimization approach that emerges from the collaborative exploration and real-time interaction between human and machine. A machine can learn the design rules for optimization of a given problem, and extrapolate design suggestions without having to search for an optimal solution in an iterative process of generation and evaluation. Accordingly, design exploration and optimization can fuse into a singular process, where through interaction of two distinct intelligences design solutions emerge. Such design methodology is of high importance for a more integrated information rich and creative design thinking. 

Looking for an emergent result of this collaborative design process, the workshop will concentrate on the generation, optimization and fabrication of freeform space frame structures with the focus on optimization of these using Finite Element Analysis and Machine Learning. The students will be introduced to OWL, a Machine Learning Platform for Grasshopper developed by Mateusz Zwierzicky. 

In the conventional methods of design optimization an iterative process of generation and evaluation is conducted in order to find an ideal solution for that certain design problem. One of the limitations of this method is computational power, as the process gets computationally cumbersome with complexity. By using Artificial Neural Networks in an intelligible manner, we can apply locally learned rules into a global system. A scalable workflow can be developed that is independent from the complexity of the given problem.

An Artificial Neural Network is only as powerful as the training set it is provided with. Therefore it is essential to understand the concept of problem abstraction and representation. The workshop will focus on formulation of design problems, by subsampling complex problems into more simple tasks to train an Artificial Neural Network to extrapolate solution patterns. 

Furthermore the workshop will be accompanied by a fabrication workflow, where the free form space frame structure developed by the students will be fabricated by a robotic bending process. In addition to optimization workflows with machine learning the students will learn to develop fabrication workflows, and prepare fabrication documents for robotic construction.

Learning objectives

At the completion of this workshop the students will:

  • Understand the fundamental concepts of machine learning
  • Learn to prepare a design problem for machine learning
  • Learn to train and validate Artificial Neural Networks
  • Learn to interactively optimize a design problem using ANNs
  • prepare fabrication documents for robotic fabrication
  • will be able to develop their own problem representations for machine learning