This thesis proposal presents a process model for design composition using a representation for the formal characteristics of design solution composition. The design process model is based on adapting previous solutions to generate new designs using derivational analogy. The model relies on a visual representation of design compositions. This representation encodes functional and form attributes of design and is used to build compositional characteristics of the design. The research proposes the use of Solution Traces (Sol-Traces) as constructs for recording and reusing the design composition path which including history, alternatives, and decisions.

Design reasoning with architectural cases has a long history in architecture. In Case Based Reasoning (CBR), an established methodology in Artificial Intelligence, the results of previous cases are applied to new situations, thus reducing the complexity of necessary reasoning and allowing a problem solver to anticipate and avoid previous mistakes. Generative CBR using derivational analogy is a powerful problem solving technique which enables new designs to be created by utilizing the generative path of prior designs. This technique is being adapted for use in developing a design assistance system for design composition in architecture. Sol-Traces are used within a CBR methodology in a computer system for design composition assistance. The research proposes a new technique for computer assisted design composition. A prototype system utilizing this approach is proposed as a proof of concept. The envisioned system provides implementation for crucial aspects of the approach: the Sol-Traces formalism and the process of design composition.

TRACE