Elsevier

Computer-Aided Design

Volume 33, Issue 14, December 2001, Pages 1111-1122
Computer-Aided Design

A case base of Case-Based Design tools for architecture

https://doi.org/10.1016/S0010-4485(01)00055-0Get rights and content

Abstract

In the 1990s, Case-Based Design (CBD) seemed an appealing approach to develop intelligent design support. Based on an alternative view of human cognition, CBD systems find new design solutions by adapting similar experiences from the past. Although several CBD applications have been built, a convincing breakthrough by these systems has yet to come. In search of reasons for this limited success, this article embarks on a critical review of the CBD approach. Its underlying cognitive model serves as a framework to analyse six CBD systems and to identify gaps in CBD research. The article focuses primarily on CBD applications for architecture, yet the findings may be relevant for other design domains as well.

Introduction

Computer-Aided Architectural Design (CAAD) has gone through many generations and philosophical perspectives. In the early and mid-1990s, a rather popular one was CBD—the application of case-based reasoning (CBR) to the task of designing [1]. In recent years, however, it has become rather quiet at the CBD front. Although the cognitive model underlying the CBD approach seems to provide a plausible explanation of how architects/designers acquire and use (design) knowledge, systems that flow from CBD research have rarely become widely used tools in offices and schools of architecture. Was CBD the umpteenth flash in the CAAD pan, or is it still a valuable path to follow?

Whereas previously we have tried to answer this question on a more theoretical level [2], this article puts some specific CBD projects under the microscope. It is so to speak a case base of CBD systems for architecture and contains six quite different cases of CBD research: Archie-II, CADRE, FABEL, IDIOM, PRECEDENTS and SEED. The six were selected because of their special concern with the domain of architectural design and because, taken together, they give a fairly good overview of the various directions in CBD research. Each case study starts with a brief introduction into the main objectives and focus of the CBD system and subsequently describes how (and whether) the different ingredients of the CBD recipe—case base (content, representation and memory organisation), retrieval and manipulation—are implemented.1 A great deal of work in this area has been published in various journals, conference proceedings and books. To our knowledge, however, this work has hardly been subject to any critique or discussion. For instance, Mary Lou Maher, Andrés Gomez de Silva Garza and Pearl Pu [5], [6] give an overview of major contributions to the field of CBD, including several systems addressed in this article, but hardly go beyond a neutral description of their features. Nevertheless, a discussion on these systems is certainly worth developing and is therefore initiated at the end of each case study. The article ends by continuing this discussion in more general terms.

Before embarking on the first case study, one may ask why we want to review CBD research in the first place. The answer is that, although a convincing breakthrough by CBD tools has yet to come, there are strong indications that encourage us not to brush the entire CBD enterprise aside. One indication is a recent experiment on the effects of using cases in architectural design, which shows that student architects effectively benefit from exposure to cases during the design process [7].

Section snippets

Case 1: Archie-II

Archie-II is a CBD aid for architects developed at the AI lab of Georgia Tech's College of Computing in collaboration with members of Tech's College of Architecture [8], [9], [10]. It descends from an earlier system called Archie, which was one of the earliest CBR applications in the domain of architectural design. Archie-II supports architects during the early conceptual stage of public building design, by providing them with interesting design cases from the past. The system focuses on case

Case 2: CADRE

CADRE—which, depending on the source, stands for CAse-based spatial Design REasoning [12], CAse-based building design through Dimensionality REduction [13] or Case Adaptation by Dimensionality Reasoning [14]—is a CBD system for preliminary building design. The project involves researchers from Architektur und CAAD (ETH Zürich), Steel Structures and the Artificial Intelligence Laboratory (both at EPF Lausanne) [15], [16], [17], [18]. The system focuses on adapting design cases to new

Case 3: FABEL

FABEL is a joint research project conducted by a consortium of six partners led by the German National Research Centre for Information Technology (GMD) [20], [21], [22], [23], [24]. Its main objective is to support architects and civil engineers in planning buildings with complex installations. Although FABEL combines several AI approaches, the central paradigm for its specification and implementation was CBR. FABEL is conceived as a collection of different tools and methods, called

Case 4: IDIOM

IDIOM—Interactive Design using Intelligent Objects and Models—is a CBD system for composing building layouts [27], [28], [29]. Coming from the same stock as CADRE, the system was developed to investigate human-computer interaction, the use of preferences, and Model-Based adaptation and combination.

Case 5: PRECEDENTS

PRECEDENTS is a CBD Aid for architecture developed by Rivka and Robert Oxman [1], [30], [31], [32].5 The system stores memorable design cases that have the status of precedents, i.e. recognised outstanding examples of a particular type or style of design, and makes these available to architecture students. The design task PRECEDENTS concentrates on is the spatial organisation of museums in the early, conceptual stage of the

Case 6: SEED

SEED—Software Environment to support Early building Design—is a hybrid design system that combines elements from both Case- and Model-Based approaches [33], [34], [35]. During the early stages of building design, the system provides support for analysis, evaluation and rapid generation of design representations. In order to do so, SEED is conceived as a collection of modules, each addressing a specific task in the design process (e.g. architectural programming, schematic layout design or 3D

Conclusion

Starting from the analysis of six CBD systems, this article has moved towards a critical review of the CBD approach. The underlying model of cognition—structure and organisation of knowledge, reasoning processes and learning—has proved to be a useful framework to inventory, analyse and discuss research in this field, and to set the agenda for the future. Inspection of this framework shows the use of specific events, case retrieval and structural adaptation as fairly well represented in the

Acknowledgements

This research is sponsored by the Fund for Scientific Research (FWO) Flanders, of which Ann Heylighen is a Postdoctoral Fellow. The authors would like to thank Professor Johan Wagemans for his invaluable suggestions for this study.

Ann Heylighen received her MS (1996) and PhD (2000) degrees in architectural engineering from the University of Leuven (K.U. Leuven). She is now a Postdoctoral Fellow of the Fund for Scientific Research Flanders in the Architecture Department of the Faculty of Engineering in Leuven. Her main areas of research are design methodology and CAAD in general, and Case-Based Design in architecture in particular. In addition, she is teaching assistant for a graduate course on CAAD.

References (35)

  • A. Heylighen et al.

    Exposure to examples: Exploring Case-Based Design in architectural education

  • E.A. Domeshek et al.

    A Case-Based Design aid for architecture

  • E.A. Domeshek et al.

    Using the points of large cases

    Artificial Intelligence for Engineering Design, Analysis and Manufacturing

    (1993)
  • Web_Archie: http://www.cc.gatech.edu/aimosaic/faculty/kolodner/archie.html (last visit 13 September...
  • B. Lawson

    Design in mind

    (1994)
  • G.N. Schmitt et al.

    Case based spatial design reasoning

  • K. Hua et al.

    Integrated case-based building design

  • Cited by (65)

    • An approach to automatic adaptation of assembly models

      2016, Computers in Industry
      Citation Excerpt :

      At present, the mainstream approaches for automatic product model adaptation fall broadly into two types: parametric design approach and combination/substitution approach. The parametric design approach mainly adopts parametric method to embed the domain knowledge into product models, and uses parameter adjustment to satisfy new design requirements [13–15]. As for the combination/substitution approach, elementary units/function units are employed to achieve target solution based on elementary unit combination or substitution [16,17].

    • Automatic shape adaptation for parametric solid models

      2015, CAD Computer Aided Design
      Citation Excerpt :

      To free the designers from these unnecessary burdens and improve the efficiency of CBD, it is very important and necessary making the case adaptation automatic. Unfortunately, although the importance of automatic case adaptation in CBD is obvious, the automatic adaptation approach is rare [2,3,5,25–30]. While some good works referred in [3,4,31,32] show that case adaptation falls broadly into two categories: adaptation in conceptual design, and adaptation in shape (geometric) design.

    View all citing articles on Scopus

    Ann Heylighen received her MS (1996) and PhD (2000) degrees in architectural engineering from the University of Leuven (K.U. Leuven). She is now a Postdoctoral Fellow of the Fund for Scientific Research Flanders in the Architecture Department of the Faculty of Engineering in Leuven. Her main areas of research are design methodology and CAAD in general, and Case-Based Design in architecture in particular. In addition, she is teaching assistant for a graduate course on CAAD.

    Herman Neuckermans is professor at the Architecture Department of the University of Leuven, where he is heading the ‘CAAD and design methodology’ research group. His research explores the use of computers in the early stages of architectural design. He teaches first year design studio, traditional construction and a course on design methodology and CAAD.

    View full text