FUZZY ADAPTIVE INTERVENTION PLAN [SUMMARY OF A DISSERTATION WORK COMPLETED IN MAY 1999].

INTRODUCTION

In this presentation we discuss the idea of a plan for the configuration of the built environment, which is generated in real time covering a wide range of scales and multiple users demands. The basic concern is in challenging traditional properties and tools that regard design as a static object, which has to be completed all at once, before the realization of space [21], [43]. Another conventional practice wants the architect to have a unique authority on the design of individual architectural objects. But the reality of the city requires a more global and collective design that completes a dialogue among individual parts and the whole structure of the city [10], [13], [33].

We will present therefore an inference construction, which generates proposals for the configuration of space by reproducing a weighed expression of users' knowledge. The plan is a Neuro-Fuzzy controller which can learn and reproduce, in general, relations of the form IF (original suggestions and intentions) …THEN (design actions and proposals), by redefining its knowledge through examples inserted from local sources. Its architecture and function are hybrid forms coming from the combination of different Neural Network and Fuzzy Controller models. The plan, which is called "Fuzzy Adaptive Intervention Plan", works in two diverse directions. The first refers to a procedure of condensing information so as to get spatial images that express multiple reasons but only one fuzzy object. This part of the plan is discussed further in detail, as it is the one on which we have concentrated our study. The second direction of the plan is focusing on the task of "defuzzication", the transformation of fuzzy proposals into clear actions and decisions for the nature, form and position of the architectural products.