Tuesday, February 08, 2005

Progress Report February 2005 (contains year 1 and year 2)

Progress Report February 2005 (contains year 1 and year 2)

I have been enrolled in the program since September 2001, and try to submit the registration to the committee in around July 2002. Unfortunately beyond my means, the research registration had only been approved in May 2003.

I spent the year 2002 doing the research background on the literatures and studying the alpha syntax program, a specific computation method to evolve branching configuration of occupational and circulation spaces. I had a vague definition of research problem surrounding the idea of relation of cell spaces and implementation socially functioning space into synthetic model after I passed the MSc at the Bartlett, and briefly tutored by Bill Hillier. Thus the literature research consequently was investigation into the work that has been done in the attempt to automate spatial layout.

I was also critically investigating a computational method called the Classifier System (CS), a machine learning system that learns syntactically simple rules from input sources such as an environment, and interested to use it as a method to create a learning alpha syntax evolution.

The result of this investigation was that although conceptually it is a relatively simple mechanism, but the whole procedure reflects as too rigidly technical, where the calculating part seen to be much more dominant than that of the artistic, explorative side. And for the kind of research problem I am interested in to, the CS probably will serve a purpose when I already got an interacting system of space and system of people running on computer before hand.

So I decided to start evaluating agent’s system, which representing occupational state in space. CECA has numerous variety of such models built so it is possible to adapt them into the project. This means the agent’s system will be a modification work, not a bottom start project. On the other side, I was looking for a way to generate the system of space that should be flexible enough to be developed into an evolutionary alpha syntax system.

By the end of 2002, I had managed to define the problem as “to find a way to generate a synthetic model of spatial configuration based on Hillier’s (1996) socially functioning of space.” These models will be based on a concept that “… a configuration of space can be influenced by, or influence, a configuration of people…”

At the end of February 2003, Paul and Christian came back from a visit to ITH and showed me a voronoi-based project called Kaisersrot[i] published in the a+t. It is a software generating urban plan which is based wholly on its community demand. After reading introductory literatures on voronoi diagram, I came to decide that it is the method for generating spatial configurations. Some features that ticked the project’s requirements are that voronoi is a free-form, divisioning of space method; it started with at least 2 points to define spaces, subsequently, and accordingly, any additional cells will associate with arbitrary points in space.

The other thing about voronoi is that it is flexible enough to develop an evolutionary alpha syntax that based on the idea of socially functioning space. The cells in voronoi, for example, can be colour coded to show the characteristics of alpha syntax structure. With a predefined allocated C cell, or the entering point as defined by Hillier, the generated voronoi diagrams can actually show occupational spaces and circulation spaces according to agent’s presence.

Thus from April to July 2003, with my voronoi’s higher language procedure in hand, Paul helped to program it in. The lack of my knowledge about program’s data structure has preventing me to code by myself, until now I am still struggling to build my programming knowledge.

Without Paul’s effort in building the voronoi, there will never be a voronoi-diagram macro for autoCAD. I look forward to learn programming properly after finish the project, so I can do the programming bits myself. Afterwards, Paul still helped around with programming to create the mechanism whereby voronoi is being fed by the presence of agents in space. This programming part has done in early 2004, with revisions here and there still going on.

I also spent 2002-2004 by studying the concepts behind agents modelling such as history of artificial life and its applications, and learn more about voronoi diagram, its origins and current applications. I also dwelled quite deep into the origins of the concept of evolutionary machine, specifically those inspired from the work of Chillean biologist, Maturana.

Maturana’s concepts such as Autopoiesis and Languaging are particularly related to the project because it contains the how of “how to know” which is essential to program-in the interaction between configuration of space and configuration of people, particularly to translate the notion of “influence” and “ influenced by” from that of Hillier’s concept.

And part of Maturana’s theory, being a biologist, concerns on the idea of organisations, the very thing I work on for the project, i.e. the configurations. It has explanations on how organisation emerges and processes involved.

I also critically study the idea of emergence from other sources, because in some agents modelling there are instances where there are emergent form of behaviours, i.e. agents have done things in the environment where these have not been programmed in by the programmer. There are 2 sources that I am interested in, Paul Cilliers and John Holland.

Paul Cilliers uses a post-structuralist approach to complexity, where he believes in the idea that a complex phenomenon needs not a single method to understand it, and it leaves all the problem of representation in the model itself. Accordingly, I long believe in architecture as a complex, not only the product and its present use, but also the process in the making and any other precede it.

John Holland’s work known as CGP shows a fit mechanism that serve a sensitive feedback to a system, where I seen the analogy of it to that of Stanford Anderson’s Problem Worrying concept. The problem of architecture is way too complex to grasp in the beginning, thus the logical approach is to work it out continuously through out.

Cilliers with his neural network projects did not come up with some recipe to program emergent phenomenon, but Holland did come up with CGP, the basic procedure he suggests, is required to start it. Holland is like other scientist, a reductionist and his CGP reflects how he looks at it as the problem to reduce phenomena and explain the emergence of it.

One particular example in Holland’s book is that emergent phenomena could be drive by a basic set of mechanisms e.g. mechanisms that is responsible in moving an electron from 1 orbit to the next. Thus, in spatial configuration and configuration of people, the mechanisms are their presence and these presences will feed to each other, resulting in the change of state of both system. The chain of mechanisms as explained by Holland will then be the generating procedure, and thus in the configurational context will be the evolution of automated layout.

In May 2004 I have given a 3 part seminars to MSc’s students, containing the background work in early literature review and the report of associate analysis of what can be observed from the model. The first part was a presentation of the research that had been done in the attempt to automate design. The second part was a presentation of the agent modelling within a mobile-spatial context, and the third part was a presentation of the voronoi diagram utilised in spatial context.

In July 2004, I gave a 3 minute presentation in the First Conference of Design Cognition, Boston MIT and presented the poster to the hall audience for 2 days. The material presented was the project with theoretical background submitted for the journal which date to publish in November this year.

With the current content and work, I am confident to draft the report this year, while will still be working on the model and learning programming from Paul. I am looking forward to publish 2 papers in peer reviewed journals and attending the next conference of Design Cognition 2006 in the Netherlands with the latest report from this project.
[i] http://www.kaisersrot.com revised 2005