A Design Assistant Architecture Based on Design Tableaux


One of the major topics in design research is the use of computational means to support knowledge based concept synthesis and creativity. While an abundance of systems has been presented in the past (based on analogy engines, case-based reasoning, genetic algorithms etc), starting with the second half of the 90s’, we can observe a progressive shift towards the exploration of agent-based systems. For example, Grecu and Brown (1996) proposed a system for the parametric design of springs. They used agents called single function agents (SFAs) that have each a unique function such as the selection of parameters, estimating their values, evaluating alternatives, criticizing or recommending. The limited functionality of these agents reinforces their interaction. Later work has introduced learning into the SFAs framework. Campbell et al. (1999) present a bottom-up approach named A-design. Various types of agents exist in the system. Configuration agents are responsible for introducing the parts they represent when and where they can into the design that is being formulated. Instantiation agents fill out these configurations with components from a catalogue. Fragmentation agents can delete inappropriate configurations or add to the catalogue configurations created in the process if they are evaluated to be satisfactory. Managing agents are in charge of the agent population; they can create or destroy agents based on their level of contribution. Gero and Reffat (2001) used multiple representations for a single agent. The agent can perceive in various ways a design, which allows changing the trajectory of design in a situated process. Reffat (2002) built on this system to produce a multi-agent version capable of creating concepts. Saunders (2001) presents curious agents that can detect novelty and originality based on Wund curves. He uses this framework for the design of art expositions to simulate the ways to maintain the public’s interest during the visit. This shift towards a Multi-Agent System paradigm in Design Support Systems research follows the general tendency of computer science striving for more and more massively distributed, autonomous systems. During the construction of design tools, to be able to cope with the fast-evolving computer science and technology, while also taking into account the particularities of design activities, one possible strategy is to build tools that make use of design theories and models. Such a strategy has the merit of having dedicated and focused solutions and also it would offer the possibility to discuss and benchmark various tools on a design theoretical level.

8 Figures and Tables

Cite this paper

@inproceedings{Hendriks2011ADA, title={A Design Assistant Architecture Based on Design Tableaux}, author={Luuk Hendriks and Akin Osman Kazakci}, year={2011} }