Breeding Objects Experiment 01 is my first experiment with configurators for mass customization of products over the Internet, conducted as part of PhD research at the Bartlett School of Graduate Studies. The experiment is a comparison of two types of configurators: “parametric” configurators and genetic algorithms configurators. In parametric configurators users have to specify the values of each of the parameters explicitly, while in genetic algorithms configurators they are not required to do so. Thanks to artificial selection, they can instead navigate through solution space by breeding objects, that is manually selecting parents of next generations.

The experiment is only about the patterns of interaction with the tools available, so you will not be asked to fill in any questionnaires. Once you start interaction with the tools, you will be able to navigate between them using links on top of each page.


Warsaw as an Emergent Structure is an initiative to acknowledge that a city is a complex system. It is also a book featuring texts from researchers and theoreticians from various fields, centered around the concept of complexity in urban environment. Ola Wasilkowska, who initiated it, aims at creating a multidisciplinary round table, aiming at developing new means of understanding the city and managing it’s growth.

Ola Wasilkowska commissioned me to write a prototype of an online platform for planning negotiations. Em_Wwa 1.0 aims to be an amalgam of top-down and bottom-up development. The municipality is able to control global parameters such as the percentage of particular functions on the site and the size of the new developments. Developers can seek places for their initiatives using a cellular automata-like algorithm, which evaluates their propositions accordingly to what has already been generated and to lighting, transportation and accessibility constraints. Citizens can engage in the process by creating their own propositions of developments and commenting those generated by others.

The site used in the EM WWA 1.0 prototype is Plac Defilad – a large square in the heart of Warsaw. The square surrounds a once controversial monument of architecture in favor of ideology – the Palace of Culture. Numerous competitions for a master plan of this square where organized to date, but it remains undeveloped as it gradually morphed from a site into a symbol.

The software was written in Processing. An animation showing EM WWA 1.0 running will be online soon.


Above are a couple of pictures showing manufacturing process of Evolve Ring Silver, commissioned and distributed by Fluid Forms and co-designed with Krystian Kwiecinski. The manufacturing process begins with a wax 3d print of a positive form. Then a rubber mold is created with the use of the print and once it is ready the actual object is cast in silver.

The entire process takes place in Austria, where Fluid Forms is based.


In general, generative design discussion is in English and efforts such as translation of Lift Architects’ Grashopper Primer to Spanish are rare. In Polish blogosphere, Przemek Jaworski is making a great effort to write a series of Processing tutorials on his blog projektowanie parametryczne.pl. There are two so far, but more is coming. First one starts from the very beginning, so someone without programing background can definitely give it a go.

tutorial 1 – Wstep do Processing (Introduction to Processing)

tutorial 2 – Krzywe Beziera (Bezier Curves)


Evolve rings where commissioned and are distributed by Fluid Forms. I have co-designed them together with Krystian Kwiecinski and they are available for purchase on Fluid Forms site since today. They where designed with an aid of an artificial selection in genetic algorithm (GA). I have developed the phenotype together with Krystian and then wrote a purpose-built GA tool for breeding, rather than designing products in Processing. We will convert the GA tool into a configurator for mass customization of the rings. It should become available on Fluid Forms site early in 2010. Below is an explanation of the roots of evolutionary approach in design, which I wrote for the purpose of the launch.

Roots of the Evolve ring generative design process

Today Darwinian model of evolution by natural selection is widely acknowledged. It assumes that a new generation of individuals within a given specie inherits properties of these individuals from the previous generation, which where the most fit to survive. Fitness to survival is always evaluated within a given environment. It is not objective, but a subject to particular conditions which the individual happened to find itself in.

On a metaphorical level, theorists has refereed to evolution of culture, societies and products that they develop. Philip Steadman’s “The Evolution of Designs” from 1979 is a great roadmap of evolutionary metaphor in reference to any kind of artifacts, including designer objects and architecture. Bashford Dean’s study of historical evolution of helmet design or Lane-Fox’s and Pitt-Rivers’ tracing of evolutionary relationships of Australian weapons are just some of the examples discussed in Steadman’s influential book.

With the dawn of computing, evolution could become an applied tool used for seeking answers to particular problems rather than a conceptual model only. Survival of the fittest has been a source of inspiration for developers of “evolutionary algorithms”, which enable navigation through vast spaces of potential solutions in any imaginable field from marketing to transportation. They also help find solutions which are a best fit for given criteria, even if these are very complex and somewhat contradictory. The art world was quick to adapt the notion of evolution and survival of the fittest, but engineering was as well. In the art world William Latham imagined artists growing artworks rather than creating them already in 1985.

Later his renowned collaboration with IBM’s Stephen Todd led to development of the software called “Mutator”. The software literally enabled Latham to breed his sculptures just like the gardener breeds plants. In this case the individual most fit to survive is the one the artist likes the most. The possibility of breeding jewelry is also mentioned in Latham’s and Todd’s contribution to Evolutionary Design by Computers edited by Peter Bentley.

This is in contrast to application of evolution in engineering, where aesthetics do not have a significant impact on the validity of the solution. What is important in this case is performance. An example is the work of dr Peter Bentley and Stephen Manos on optimization of Holey optical fibres. These fibres are able to transmit light on large distances and their capacities are greatly enhanced by small air holes running through their entire length. Question was what shape, size and number of the wholes provides a fibre with best transmission capacity? Looking for an answer with evolutionary computation likely took millions of solutions to develop, but as the process was automated the time lapse was not so big and the results where rewarding.

Evolve ring brings evolution to the design world, not as a metaphor, but as a design tool. In design, aesthetics plays a crucial role, but the design has to be valid from a manufacturing stand point as well. This is why an algorithm used by Michal Piasecki and Krystian Kwiecinski is an amalgam of the GA from the art world and from the engineering world. Michal and Krystian developed a parametric phenotype first. Later Michal unleashed the process of evolution of jewelry and guided it towards the designs, which became the Evolve rings designer’s series from Fluid Forms. These pieces are breed, not designed.


Westiwal

Przemek Jaworski (who runs the projektowanie parametryczne.pl blog) and I where invited to speak about computational design at the “Westival” – architectural festival in Szczecin.

Our talk will take place on 17:00, 23rd of Nov in “Willa Lentza”, al. Wojska Polskiego 84. Program of the event is available here (polish). Please join us if you are in Szczecin at the time.

We will begin with a bit of history of computational design (John Frazer will certainly be mentioned). An introduction to computational tools such as parametric modeling, scripting and programing will follow. Later we will speak about optimization with genetic algorithms and particle-spring systems. Finally we share out experiences from Adaptive Architecture and Computation course at the Bartlett.


ACDSeePrint Job

John Frazer’s “An Evolutionary Architecture” from 1995 is an obligatory read for anyone interested in the amalgam of evolutionary computation and design. In other words anyone interested in the potential of “breeding designs” using evolutionary algorithms (genetic algorithms belong to this category). I was very happy to see it available online for free from the AA website.


NURBS surface Processing

Couple of quick impressions from a Processing sketch, which I’m currently developing. The sketch is going to become a basis of my first PhD experiment: comparison of a “parametric” interface and a “genetic algorithm” interface of a configurator for mass customization of furniture. The experiment will actually be conducted online, so more details will follow. For now I will only say that in order to compare two interfaces I need a parametric definition of an object, which can be manufactured with CAM. I have decided to go for a NURBS surface with control points wrapped in both U and V dimensions.

Processing sketch was developed based on Alasdair’s Turner’s NURBS surface sketch.

Before moving to Processing I did a couple of quick sketches in Generative Components:

NURBS surface GC


There where three sessions on MCPC devoted entirely to configurators. Before moving on to discuss them, let’s answer a question: “what is a configurator?”. The definition provided by Configurators Database is a good answer:

“Simply put, a configurator is a software application for designing products exactly matching customers’ individual needs.”

B2C online product configurators are of my particular interests. These ones are the tools of communication between producers and customers (or users) by allowing the later ones to customize products of the former ones. The whole process takes place over the internet.

De-facto standards in the design of configurators

Clarissa Streichsbier’s talk on “Identification of De-Facto Standards for Designing the User Interfaces for Web Based B2C Product Configurators” was one of the most appealing ones. Streichsbier (who works with cyLEDGE) looked at similarities of layout in 126 different online product configurators from 3 different industries: electronics, apparel and automotive. The Configurator Database was her source of links to the online customization tools. Sean Hanna and me have used this database as well in order to access most of the configurators discussed in our review.

The findings of Streichsbier’s review are very consistent. Standarts can be found in designs of configurators from the same industry. For example: electronic industry seems to pay little attention to the actual picture of the product. Instead it devotes most of the screen space for the customization menu. Automobile industry’s approach is the opposite. Product visualization in this case occupies majority of the screen space. It additionally seems that there are no clear standards across the industries. This means that if an experimental research was about to be conducted on online product configurators, these de-facto standards would definitely have to be taken into consideration in the design of the experiment.

“Open Toolkit”

In a later session Günter Dressel’s and Wolfgang Frühwirt’s, also from cyLEDGE, launched the project called “Open Toolkit”. Their aim is to create an open-source tool for development of configurators. I’m looking forward to seeing how this will develop and hope to contribute as well.


MCPC logo

I have recently attended Mass Customization and Personalization conference in Helsinki (MCPC 2009), where I’ve presented the Review of B2C Online Product Configurators. There where couple of really interesting issues raised during these intensive two days and I would like to begin with discussion of the term “personalization”. Posts on automation in architecture as well as B2C product configurators will follow.

Personalization as an umbrella term

Diversity of meanings applied to the term “personalization” was quite astonishing. For example: Bruce Kasanoff used it to name nearly any kind of mass customization. His examples ranged from products up till body customization with electronic devices.

Personalization as an process in which user is passive with respect to product attributes

Sunnika and Brage’s literature review provided an interesting insight on the distinction between “mass customization” and “personalization”. Their work is an automated literature review, facilitated by text mining algorithms. They have found that “mass customization” referees more often to tangibles, while “personalization” is rather used in the context of “intangibles”. In other words mass customization is likely to be about products, while personalization is associated with information.

This goes in the same line with what Sean Hanna and I proposed in our review. We considered “mass customization” as the process in which user is active with respect to product attributes, while we referred to “personalization” as a process in which user is passive with respect to product attributes. We discussed recommender systems the examples of personalization.

Another interesting input from Sunnika and Brage is that exactly the same distinction in computer science literature was described as a difference between “adaptability” and “adaptivity”. Where “adaptability” is a process in which the user is active, while adaptivity is an automated process.