Environment and Planning B: Urban Analytics and City Science

In the first part of this paper a view is presented of what environmental planning is and how it can be implemented, providing in support a brief review of key literature on planning theory and practice. The review points to limited success in adapting mathematical programming techniques from operations research to the making of environmental planning decisions.

Computer scientists have long been interested in developing computer programs that solve problems and make decisions in the way that humans do. Over the last ten years most of this ‘artificial intelligence’ research has concentrated on knowledge-based systems and one variety of these systems, expert systems, has now been developed to the stage where it can be successfully applied to practical problems. In the second part of the paper the working of knowledge-based systems is briefly described, and the features that make them promising for planning purposes are introduced. These features can be summarized as (1) the handling of qualitative information, (2) the use of uncertain and incomplete information, (3) ease of modifying information on which decisions are based, (4) an ability to explain solutions and decisions, and (5) use of natural language for communication.

Different types of expert systems have been developed for such diverse tasks as design, classification, selection, and monitoring; tasks typically undertaken in environmental planning. In the third part of the paper ways in which these types of systems can be applied to the different stages of the planning task are suggested. Although it is too early to judge the likely success of these systems in practice, it is clear that they have the potential to overcome some of the deficiencies of earlier computerized decision aids.

Although knowledge-based systems can be shown to be of assistance to designers in a number of areas like diagnosis of faults, testing for conformity to regulations, selection of elements and components, and so on, there are problems in their more general design application. Some of the difficulties that arise are examined together with possibilities for their resolution. Parts of the discussion centre around the problems of maintaining semantic integrity in a developing design. A possible design system based on the concept of the design knowledge manipulation environment is introduced.

FireCode is an expert advisory system for assisting building designers to check their plans against the requirements of a fire safety code. This application of expert-systems methodology is rather unusual because much of the knowledge is already available in written form, reducing the effort required for development. However, expert assistance was still required to handle a number of problems in the interpretation of the written code, such as incomplete and ambiguous requirements. These problems are likely to be met during the development of any expert system based on regulations or codes of practice. A strong parallel exists between the development of an expert system and the development of a code of practice; they are both concerned with the formalisation of knowledge. An important conclusion of this work is that the parallel development of an expert system will assist in the formulation of a code of practice.

Grammars of shapes and designs are rule-based models for the generation of members of a class of designs. The applicability of the formalism as a basis for generative expert systems for construction details in computer-aided design is discussed. A demonstration system is presented.

Several years' experience in building substantial prototype expert systems (involving several thousand lines of Prolog), mostly for engineering design codes using graphics, has led to the conclusion that Prolog is a very awkward language for representing several types of knowledge that are common in design codes. These include rules that are naturally written as mathematical expressions, operations that require iteration over the members of a list, handling of complex tables and arrays, and graphical user interfaces.

An expert-system shell is now being designed to overcome these problems. The shell will use a declarative language based on Prolog, but with extensions to handle the above problems and to facilitate explanation of the reasoning process. It will be prototyped in Prolog, and then implemented in C to achieve greater efficiency.

The knowledge-representation problems that have inspired the development of this shell are described, and preliminary proposals for the design of features to overcome them are presented.

Increasing specialisation and the growth of automated advice-delivery systems are creating new problems in legal responsibility and ethical behaviour. Engineering, planning, legal, and medical workers can expect early encounters with these difficulties, which are essentially concerned with a new interpretation of ‘due care’ and of ‘professional liability’. The precipitating factor in this debate is the emergence of usable ‘expert’ systems, which embody judgmental and operational knowledge, and are often designed to mimic the behaviour (if not the public pronouncements) of acknowledged experts in the field. The task of the knowledge engineer and of the professional worker using—or expecting others to use—such automated advisory systems raises ethical problems both for individuals and for professional and learned societies. Some of these are considered in this paper with special reference to the arms-length delivery of engineering, planning, and regulatory advice.

Shape grammars are specified that generate houses in the Queen Anne style which dominated domestic architecture in the United States of America in the 1880s; examples are used which are typical for Pittsburgh's historic Shadyside district. Separate grammars are given for the generation of plans and for the articulation of plans in three dimensions. Both grammars emphasize aspects of geometry and overall design and explain how individual parts and features are related to each other.