Architectural Design Task Research Articles

Parametric Development of Problem Descriptions

This paper addresses the development of parametric models in contemporary architectural practice. A parametric model can be regarded as a representation of a solution space and in order to structure this, a description of the problem is required. Architectural design tasks are typically ill structured, the goals may not be defined and the means unknown. Moving from an incomplete problem description to a functional parametric model is a difficult task. In this paper the aim is to demonstrate that through a combination of knowledge acquisition and capture a parametric model can develop from an incomplete problem description. This demonstration draws on existing strands of design theory which are then used to outline a theoretical framework. This framework is then used to examine a case study of a live project and practical examples of the described theory in action are given. The practical observations are the result of a case study involving the author as a participant and observer working with Populous to develop a cladding geometry solution for Lansdowne Road Stadium in Dublin (now know as the AVIVA STADIUM).

EvoArch: An evolutionary algorithm for architectural layout design

The architectural layout design problem, which is concerned with the finding of the best adjacencies between functional spaces among many possible ones under given constraints, can be formulated as a combinatorial optimization problem and can be solved with an Evolutionary Algorithm (EA). We present functional spaces and their adjacencies in form of graphs and propose an EA called EvoArch that works with a graph-encoding scheme. EvoArch encodes topological configuration in the adjacency matrices of the graphs that they represent and its reproduction operators operate on these adjacency matrices. In order to explore the large search space of graph topologies, these reproduction operators are designed to be unbiased so that all nodes in a graph have equal chances of being selected to be swapped or mutated. To evaluate the fitness of a graph, EvoArch makes use of a fitness function that takes into consideration preferences for adjacencies between different functional spaces, budget and other design constraints. By means of different experiments, we show that EvoArch can be a very useful tool for architectural layout design tasks.

Knowledge Acquisition in Parametric Model Development

This paper addresses the development of parametric models in contemporary architectural practice. A parametric model can be regarded as a representation of a solution space and in order to structure this, a description of the problem is required. Architectural design tasks are typically ill structured, the goals may not be defined and the means unknown. Moving from an incomplete problem description to a functional parametric model is a difficult task. This paper aims to demonstrate that through a combination of knowledge acquisition and capture a parametric model can develop from an incomplete problem description. This demonstration draws on existing strands of design theory which are then used to outline a theoretical framework. This framework is then used to examine a case study of a live project and practical examples of the described theory in action are given. The practical observations are the result of a case study involving the author as a participant and observer working with HOK Sport to develop a cladding geometry solution for Lansdowne Road Stadium in Dublin.

Efficient Case-Based Structure Generation for Design Support

This paper describes a general approach to support case-based structure generation named Conceptual Analogy (Borner 1997). The approach can be used to support design tasks in domains that do not allow the acquisition of a complete and consistent set of constraints or rules but that do provide a larger set of past experiences. The experiences – also called cases – may resemble for example CAD data of room layouts or pipe systems. Thus, cases represent solutions to particular problems without an explicit, a priori separation of problem and solution variables. The paper starts with a description and formalization of the case-based structure generation task and a discussion of why standard case-based reasoning and other approaches are not applicable. Then, the new approach of Conceptual Analogy (CA) is introduced. CA applies conceptual clustering to organize cases represented by graphs into hierarchical classes of structurally similar cases. These case classes are then represented by concepts. Given a problem, i.e., a partial solution, the hierarchy of concepts is searched for applicable concepts, i.e., concepts that allow the generation of at least one solution to the given problem. Applicable concepts are used to generate a set of solutions that can be ordered according to their quality. Properties of the approach as well as complexity results are presented. An architectural design domain and task where the approach has been applied successfully, is used for illustration and for practical evaluation. Finally, the approach and its implementation are compared to two systems that aim at the support of similar design tasks. The paper concludes with an assessment of the future direction of this research.

Learning algorithms based on linearization

The aim of this article is to investigate a mechanical description of learning. A framework for local and simple learning algorithms based on interpreting a neural network as a set of configuration constraints is proposed. For any architectural design and learning task, unsupervised and supervised algorithms can be derived, optionally using unconstrained and hidden neurons. Unlike algorithms based on the gradient in weight space, the proposed tangential correlation (TC) algorithms move along the gradient in state space. This results in optimal scaling properties and simple expressions for the weight updates. The number of synapses is much larger than the number of neurons. A constraint for neural states does not impose a unique constraint for synaptic weights. Which weights to assign credit to can be selected from a parametrization of all weight changes equivalently satisfying the state constraints. At the heart of the parametrization are minimal weight changes. Two supervised algorithms (differing by their parametrizations) operating on a three-layer perceptron are compared with standard backpropagation. The successful training of fixed points of recurrent networks is demonstrated. The unsupervised learning of oscillations with variable frequencies is performed on standard and more sophisticated recurrent networks. The results presented here can be useful both for the analysis and for the synthesis of learning algorithms.

Learning algorithms based on linearization

The aim of this article is to investigate a mechanical description of learning. A framework for local and simple learning algorithms based on interpreting a neural network as a set of configuration constraints is proposed. For any architectural design and learning task, unsupervised and supervised algorithms can be derived, optionally using unconstrained and hidden neurons. Unlike algorithms based on the gradient in weight space, the proposed tangential correlation (TC) algorithms move along the gradient in state space. This results in optimal scaling properties and simple expressions for the weight updates. The number of synapses is much larger than the number of neurons. A constraint for neural states does not impose a unique constraint for synaptic weights. Which weights to assign credit to can be selected from a parametrization of all weight changes equivalently satisfying the state constraints. At the heart of the parametrization are minimal weight changes. Two supervised algorithms (differing by their parametrizations) operating on a three-layer perceptron are compared with standard backpropagation. The successful training of fixed points of recurrent networks is demonstrated. The unsupervised learning of oscillations with variable frequencies is performed on standard and more sophisticated recurrent networks. The results presented here can be useful both for the analysis and for the synthesis of learning algorithms.

What do architects and students perceive in their design sketches? A protocol analysis

The present research aims at examining what information architects think of and read off from their own freehand sketches, and at revealing how they perceptually interact with and benefit from sketches. We explored this in a protocol analysis of retrospective reports; each participant worked on an architectural design task while drawing freehand sketches and later reported what she/he had been thinking of during the design task. This research lies within the scope of examinations of why freehand sketches as external representation are essential for crystallizing design ideas in early design processes.

A study of human performance in computer‐aided architectural design

This paper describes the development and application of a cognitively‐based performance methodology for assessing human performance on computer‐aided architectural design (CAAD) tasks. Two CAAD tasks were employed that were hypothesized to be different in terms of the underlying cognitive processes required for these tasks to be performed. Methods of manipulating task complexity within each of these tasks were then developed. Six architectural graduate students were trained on a commercially available CAAD system. Each student performed the two experimental design tasks at one of three levels of complexity. The data collected included protocols, video recordings of the computer screen, and an interactive script (time‐stamped record of every command input and the computers textual response). Performance measures and methods of analysis were developed which reflected the cognitive processes used by the human during design (including problem‐solving techniques, planning times, heuristics employed, etc.) and ...