This paper describes SOJA, a scheduling expert system for planning a workshop’s daily activities. The paper consists of two parts. The first part describes the conceptual data representations and algorithms employed in the selection and constraint-directed scheduling phases of the problem solving process. The second part describes a LISP-based implementation of the system.

The first part of the paper is well written and easy to read. Two interesting points that come out are: (1) the casting of classical Operations Research (OR) techniques and methods into a pattern-directed inference mechanism, and (2) the combining of these with shop-floor heuristics about scheduling. The general scheduling problem has long been a classical domain of application for OR techniques and methods, and this paper, like many others, fails to elucidate why AI/Expert Systems techniques build better scheduling systems.

The second part of the paper presents an implementation of the ideas described in the first part. However, reading this part is an uphill task because the implementation details are freely intermingled with the author’s preoccupation with how the implementation should be “viewed.” There are ambiguous references to the employed representations as “frames” and the maintenance of matched partial instances of rules as a way around the “frame problem.” Other than this confusion with the “frame” issue, the paper is OK.

Those involved in the construction of knowledge-based systems for manufacturing and industrial automation may want to browse through this paper.