There exists a large body of Artificial Intelligence (AI) research on generating plans, i.e. linear or non-linear sequences of actions, to transform an initial world state to some desired goal state. However, much of the planning research to date has been complicated, ill-understood, and unclear. Only a few of the developers of these planners have provided a thorough description of their research products, and those descriptions that exist are usually unrealistically favorable since the range of applications for which the systems are tested is limited to those for which they were developed. As a result, it is difficult to evaluate these planners and to choose the best planner for a different domain. To make a planner applicable to different planning problems, it should be domain independent. However, one needs to know the circumstances under which a general planner works so that one can determine its suitability for a specific domain.

This paper presents criteria for evaluating AI planners; these criteria fall into three categories: (1) performance issues, (2) representational issues, and (3) communication issues. This paper also assesses four non-linear AI planners (NOAH, NONLIN, SIPE and TWEAK) based on a study of the published literature and on communication (via electronic mail, meetings and correspondence) with their developers.

A generalized approach to fast interpretation of objects and their features has so far eluded researchers. In manufacturing, this interpretation can be approached from the vision point of view or from the CAD data perspective. Presently, CAD systems are widely used in several aspects of manufacturing production. It is therefore more efficient to use CAD data for object reasoning in manufacturing, especially when systems will eventually be data driven. Components can be modelled on a CAD system using various modelling techniques and the representation of their geometric information is still CAD system dependent. However, the advent of the Initial Graphics Exchange Specification (IGES) now makes it possible to represent CAD data in a neutral and standard manner.

This paper describes a scheme for recognizing and representing features for CAD data extracted using the IGES interface. The concepts developed are based on graph-based feature representation, where features are represented by a set of faces as well as their topological adjacency.

Strategies for classifying features and methods of decomposing a complicated feature into several simpler features for recognition purposes are discussed.

We show that the usual notion of constraint propagation is but one of a number of similar inferences useful in quantitative reasoning about physical objects. These inferences are expressed formally as rules for the propagation of ‘labeled intervals’ through equations. We prove the rules' correctness and illustrate their utility for reasoning about objects (such as motors or transmissions) which assume a continuum of different states. The inferences are the basis of a ‘mechanical design compiler’, which has correctly produced detailed designs from ‘high level’ descriptions for a variety of power transmission and temperature sensing systems.

A blackboard based intelligent control system has been developed for a family of complex non-equilibrium materials processes. The system is being tested in the laboratory for control of a particular high risk, high value-added step in the manufacture of carbon-carbon composites. The system uses knowledge based methods in several fundamental ways to fill gaps left by control theory and process models. Most notable of these are (1) inferring from indirect measurements and history the process state at multiple, changing levels of abstraction, (2) anticipating problems and planning actions to reach goal (end of process) states, (3) selecting, executing and interpreting approximate models to predict process progression and (4) changing control objectives as the physical situation changes. The system has been demonstrated to substantially reduce processing time.