A design process has two major phases, conceptual design and detailed design. The major emphasis has been on the detailed design phase, such as VLSI chip layout and fabrication, structural design, and computer-aided drafting. The earlier phases of conceptual design are where problems are identified, functions and specifications are developed, and appropriate solutions are generated through a combination of basic building blocks. Conceptual design involves a mix of numeric and symbolic reasoning.

This text concerns innovative conceptual design, which can be described as solving a known problem in a new way. In innovative design we have no prior knowledge about the shape of the solution, so we cannot predetermine a set of steps that will result in a solution. The design criteria are also likely to change during the process, and multiple design objectives can exist and must be optimized simultaneously.

The author’s approach to innovative design is exploration, evaluation, and adaptation. The exploration mode has the ability to generate alternatives that break away from a normal solution and its constraints. These alternative solutions are then evaluated for appropriateness. In the adaptation mode, prior knowledge is used to adapt faulty designs. Faulty designs occur because an interesting solution may relax the criteria in the exploration mode. If this occurs, the designer must find some way to compensate for this relaxation. Criteria emergence occurs if a precedent that satisfies some criterion not present in the original design problem is used.

The material is organized around a computer program, called CYCLOPS, that implements an exploration and adaptation mode. CYCLOPS also has a normal search mode, which performs a tree-like search of design alternatives. This mode will generate a design that satisfies all the given criteria and constraints, provided that one exists.

The ideas in the text are presented with the aid of a landscape design example. The chapters detail the three major areas of design representation, design exploration, and design adaptation. Chapters 3 and 4 are about design representation and search. Chapter 5 discusses design exploration and the emergence of new criteria. Chapter 6 is concerned with precedence-based adaptation. Chapter 7 gives the overall architecture of CYCLOPS and states how the components fit into a system. Chapters 8 and 9 discuss the relationships between this material and other research, applications, and future research in engineering optimization.

The main concepts of this book can be found in several disciplines. The consistent labeling optimization problem solving methodology is an integer programming problem with generalized constraints. The demand pasting technique from artificial intelligence is used for its ability to employ multiple precedents or parts of precedents to solve a problem. Finally, the idea of exploring design alternatives by criteria relaxation could be a component of design automation.