MSc Adaptive Architecture & Computation

the Board of Examiners for the MSc Built Environment (who met on 18 Oct 2013) recommended Stamatis Psarras to receive an MSc Pass degree with Distinction.


Over the years, architecture has found ways to benefit from the use

of computation tools, either to create interesting forms or to optimize

current designs. The optimization can happen in various levels of the

build environment; one of the most challenging is spatial analysis. This

dissertation is cantered on spatial analysis in the build environment,

particularly using agent simulation. The aim is to find the correlation

between the visual information from the perspective of the agents and

their behaviour in space. This process involves using an Artificial Neural

Network (ANN) that was trained to make accurate Agent Simulation.

The training was done using a variety of hypothetical paths and

different scenarios that represent observable data from a given space.

This methodology showed that it is possible to train agents to behave

in a particular way, given only the desirable path and the context. The

ANN made possible the simulation of behaviours beyond the scope of

current analysis tools by taking advantage of all the visual information.

Furthermore, new methods were proposed to visually exhibit the

capabilities of the ANN and derive conclusions on the particular spatial

elements that dictate behaviour. As a result this tool can be used to test

different spatial configurations at a much finer level.