Design Tooling - Design Machines

Return to front page

Information

Introduction
This module engages the design of "things that design" and explores the relationship between designers (or design mechanisms) and designs. Design is often discussed as a multi-faceted process. As such, design tools have often been developed to address isolated issues in design (e.g. analysis, generation, fabrication, evaluation). All of these systems segment the design process into isolated parts and consign coherent design integration to the intuitive devices of human designers. While this is often desirable, it is also worthwhile to explore completely autonomous design systems which perceive and act independently of human intervention. In this class of systems, the intuitive gaps must be explicitly filled and the integration of related issues must be explicitly resolved. Students will be provoked to remove themselves from direct design and focus on stating direct connections between perception and design outcome. Students will be shown how to develop simple autonomous design machines. These machines will attempt to mimic the behaviors of living creatures which build nests, hives, and mounds in complex physical environments.

Background

The 1981 paper "Design Machines" by George Stiny and L. March outlined the development of an autonomous system for creating designs. Their framework divides the design process into four mechanisms: Receptor, Effector, Language of Designs, and Design Theory. The receptor creates representations of external conditions and the effector stimulates external processes or artifacts from designs. According to Stiny and March, Receptors can be sensors or traditional input devices like a keyboard or a mouse. Effectors might be Printers, CNC machines, and even Robots. Together the receptor and effector determine the design context. The Language of Designs gives an account of the formal parameters of designs. Finally, the Design Theory describes the correspondence between the Language and the Context.

Weaver Bird

Implimentation

Introduction
The following section outlines a few simple techniques to get students started on perceiving and interpreting an environment through a computational lens. According to Stiny and March, perception is the responsibility of the receptor mechanism. The receptor and effector, taken together make up the design context or the interface between the design machine and the outside world.

Images

Image Types:

Camera Images: Produced using a lens and a light sensitive receptor.

Transmission Images: Produced when light is shone through an object. (ex. X-ray)

Sonic Images: Reflection of Sound waves off an object. (ex. Medical Ultrsound)

Radar Images: Tradiitonal Radar screen.

Pressure Maps: Information from a grid of pressure sensors.

Range Images: Matrix of distances to different objects in a space. (ex. Sonar)

Pixel-Based Approach

2D Scaned images and digital photographs can be used to import information about context into a computational platform. However, images can be produced with other means. These imported images become digital objects when they are translated into a grid of colored pixels. All image recognition and manipulation algorithms rely on procedures which examine and manipulate images at a pixel level.

Getting Started: A good place to start is with operations on Bi-Level Images. Images containing only black pixels and white pixels are the easiest to interpret and manipulate. Thresholding is a means of converting a gray level or color image to a BI-level image. One of the ways to define objects in an image is through connectivity. The 'seed' method can identify a bounded object from a single pixel seed. The algorithm checks the neighbors of the seed and the neighbors' neighbors until it has found all the boundaries of the object. Once objects are defined in a image they can be interrogated through geometric operations such as area, perimeter, etc. Through a method called erosion, "extra" pixels can be removed from an image in order to produce cleaner images in which objects can be picked out more readily.

Reference: J.R. Parker, Practical Computer Vision Using C. John Wiley & Sons, New York 1949.
Source Code

Shape-Based Approach: Shape Grammars, developed by G.Sting and J.Gips, stand as a critique of the vast majority of computational vision systems. Shape Grammars can be aligned with gesalt theories of perception in which compositions are more complex than the sum of their consitituent parts. According to Shape Grammarians, images are not composed of pixels. In fact the structure/complexity of an image is a function of the operations performed on it. Apart from its relience on shape, Shape Grammars does not adhere to any "fixed" means of decompositing images.

Refernce: http://www.shapegrammar.org

(*) Topics for Discussion

Many theorists, including Stiny, have argued that much of creative behavior is embedded in perception. "New" A.I. also claims that perception is inseparable from reasoning. Can designers benefit from discussing and implementing "ways of seeing" explicitly? How might new means of perception enabled by technology change the way that architects see their own role as designers of environments?

(**)Assignment

Design a mode of perception which embodies desirable biases. Use this perceptive mechanism to interpret various artifacts and environments.

Shape Grammars

How many ways can you decompose this shape?

Pixelization
Building up images from descrete units

Camera Image

Radar Image

Transmission Image

Range Image

Computer Vision using a WebCam by Josh Nimoy et al.

"What is Computer vision? Realtime video input digitally computed such that intelligent assertions can be made by interactive systems about people and things. Popular techniques in new media arts and sciences include the ability to detect movement and presence in spaces, appearance of objects or people, how many of them are there, which way it's facing, and edge path vectors. Myron brings computer vision to a growing number of interactive media development platforms, allowing cameras connected to your computer to control just about anything. This software aims to make computer vision easy and inexpensive for the people! Currently, it has more "tracking" functionality than other plugins with similar aim. "

References: http://webcamxtra.sourceforge.net/

Web Cam Still

Pixelized

Bubblized

Vector Trace

Tracking

Microworlds

Introduction
Microworlds offer constrained environments whose very nature helps users to develop an understanding of complex causal relationships. The goal of microworlds is, simply stated, "to develop new external systems of representation that foster more effective learning and problem solving." (Goldin 1991) Theories of Piagetian learning have been one of the main influences in the development of microworlds. Piaget's theory of constructivism asserts that people construct knowledge about the world through experience. The first developers of microworlds held that one's construction of knowledge is aided by the process of making external or shareable artifacts. Microworlds can help architects learn about the implicit logic of organizational and material systems in buildings by constructing the logics of those systems explicitly.

Educational Examples
Logo (Papert), the Sim Series (SimCity, SimLife, The Sims), and StarLogo (Resnick) are computer programs that help people make sense of the world by making things in the computer. Logo was developed at MIT by Seymour Papert, a well known mathematician and advocate for computers in education. Logo is a virtual environment intended to teach children mathematics by allowing them to draw pictures through numerical instructions to a virtual drawing agent, the turtle. Children playing within the world of Logo learn to understand the nature of geometrical figures. (ex. a square has four equal sides and four equal angles) StarLogo is an elementary programming language designed by Mitchell Resnick for people with no programming experience. "With StarLogo, people can write rules for thousands of graphic creatures on the computer screen, then observe the group-level behaviors that emerge from the interactions."(Resnick 1994)

References:
Edwards, L. Microworlds as Representations.
Goldin, G. (1991) The IGPME working group on representations. In F.Furinghetti (ed) Proceedings of the XV Conference of the International Group for the Psychology of Mathematics Education vol. 1, p. xxvii: Assisi, Italy.
Papert, S. (1980). Mindstorms: Children, Computers, and Powerful Ideas. New York: Basic Books.
Resnick, M. (1994) "Learning About Life." Artificial Life, vol. 1, no. 1-2.

Developed by John von Neumann(1903-1957). "Cellular automata are discrete dynamical systems whose behavior is completely specified in terms of a local relation. A cellular automaton can be thought of as a stylized universe. Space is represented by a uniform grid, with each cell containing a few bits of data; time advances in discrete steps and the laws of the "universe" are expressed in, say, a small lookup table, through which at each step each cell computes its new state from that of its close neighbors. Thus, the system's laws are local and uniform." (Brunel University Artificial Intelligence Site)

Example by Simon Greenwold
http://www.architecture.yale.edu/872a/processingExamples/CA3D_Template/index.html

Suggested Assignments
1) Design a rule set for the Cellular Automata which deals provocatively with figure-ground relationships.
2) Construct a critique of the voxel approach to designing in 3D.

Simon Greenwold 2003

Introduction
Braitenburg Vehicles* demonstrate how organization can emerge out of the interaction of a set of simple machines operating without centralized control. These examples build on code originally written by Simon Greenwold. Braitenburg vehicles are implemented here as computational systems in which simulated sensors and motors are linked to produce behavior. These vehicles have left and right light sensors and left and right motors. All these mechanisms are defined independently but they can be linked in different ways to imbue vehicles with different behaviors. Links between sensors and motors can follow one of three underlying schemes:

(1) left sensor to left motor / right sensor to right motor
(2) left sensor to right motor / right sensor to left motor
(3) both sensors to both motors

Vehicles respond to the environment
In this example the mechanisms are wired according to scheme (2). The speed of each motor is directly proportional to the amount of light detected by its corresponding sensor. In the case of scheme (2) crossed sensor/motor links cause the vehicles to turn toward light spots. The vehicles have been placed on a black and white photograph depicting some diffuse shadows. As the vehicles traverse the 2D space of the photograph, they appear to be attracted to or to flee from different features.

*Braitenburg, Valentino. Vehicles. Experiments in Synthetic Psychology. MIT Press, Cambridge 1986

Sticky Environments
This is a 2d example in which the imagemap is populated by small bright "sticky" elements which can attach to vehicles and other elements within certain proximity. Over time, elements are rearranged by these interactions. The vehicles also leave color trails as they move. These trails can be reinforced by the trails of other vehicles. Both the trails left by the vehicles and the arrangement of sticky elements change the feature space of the imagemap. This creates a feedback loop, influencing the future activity of vehicles.

Yanni Loukissas 2004

Anatomy of a Braitenburg Vehicle

2D Vehicles, 3D Trails
The intent behind this example is to develop an interface for students to explore design machines in a 3D environment. The 3D vehicles in this example can be used to simulate environmental forces or to search for emergent organizational phenomena. In the applet below, Braitenburg vehicles move over a 2d imagemap collecting information about light and dark spots. This information is used to construct forms in 3d. In the more advanced examples. Information from the 3d form is projected back onto the source imagemap. This is example is a basic extrusion of the standard 2d Braitenburg model according to brightness levels.

Yanni Loukissas 2004

Preliminary Controls: Use "Shift" , "x" and "z" with the mouse to pan, zoom and rotate.

Shadow Constructor
The vehicles in this example builds surfaces instead of trails. They are still guided by a source imagemap. The constructed surfaces cast shaddows on the imagemap. This results in a feedback loop which augments the behavior of vehicles.

Yanni Loukissas 2004

Preliminary Controls: Hold down "d" to draw. Press "q" to start the vehicles. Use "Shift" , "x" and "z" with the mouse to pan, zoom and rotate.

Advanced Shadow Constuctor
This example builds on the previous two and add the ability to have multiple imagemaps and more control over shadows.Users have the option to select one or more overlaping imagemaps as a starting point. Vehicles can read many overlaid maps. Users can also manually place light and dark sections on the imagemap. This is just a starting point for the exploration of multiple level microworld investigations using light and shadow as motivating parameters. The next step is to add a new class of vehicles which create "activity areas" in which augment and respond to the distribution of light and shadow. Design students playing with multi-level microworlds like this have the opportunity to explore rule-based systems which are sensitive to multiple dimensions of contextual information.

Version1

Version2

Yanni Loukissas 2004

Preliminary Controls: Hold down "d" to draw. Press "q" to start the vehicles. Use "Shift" , "x" and "z" with the mouse to pan, zoom and rotate.

Topics for Discussion
By working with decentralized computational systems, students of architecture can explore how designs might be developed in a "bottom up" fashion. Students might also come to understand the interdependence of design behavior (perception and action) and context. With respect 3d, How does it change the stakes for the use of computation in design? What additional kinds of investigations does a third dimension allow?

Suggested Assignments

1) Design an environment to accommodate a programmed vehicle (ex. painting with light)

2) Design a vehicle which will build a predetermined structure in response to a given environment.

Design Games

In this educational scenario, design machines will be situated as agents within the context of abstract design games. This strategy builds on research conducted in the 80's at the MIT department of architecture in which architectural design was explored through the metaphor of a game (Habraken 1987). The games developed at that time were entitled 'Concept Design Games.' They attempted to highlight some of the vital characteristics of the design activity by focusing on the interaction of designers. This scenario proposes the development of systems which can exhibit visual reasoning of the nature required to play 'Concept Design Games.' The initial game to be explored using architecture design machines is a variation of the silent game, developed by Habraken. In the silent game, designer/players build elaborate visual compositions through the use of patterns. This game highlights the implicit understandings that develop between designers in making and projecting patterns through visual compositions.

Topics for Discussion

As human designers/players attempt to anticipate the behavior of the machine they become engaged in thinking about how design happens. The interaction between human and machine designer/players can be a controlled and informative opportunity for teachers and students alike to reflect on intuitive and mechanistic approaches to design.

Suggested Assignment

1) Play the silent game with a design machine and discuss the results

2) Develop a new concept design game and program a design machine to play it.

Yanni Loukissas 2004

Return to front page