EXPERT SYSTEM TECHNOLOGY IS GROWING RAPIDLY.

EXPERT SYSTEM TECHNOLOGY IS GROWING RAPIDLY. COMMERCIALLY AVAILABLE PACKAGES (GENERAL PURPOSE INFERENCE ENGINES, KNOWLEDGE/DATA REPRESENTATION SCHEMES AND DEVELOPMENT ENVIRONMENTS) ARE MULTIPLYING. CUSTOM REAL-TIME EXPERT SYSTEMS ARE NOW BEING FIELDED. REAL-TIME FEATURES ARE BEING ADDED TO GENERAL PURPOSE PACKAGES. MAJOR STUMBLING BLOCKS STILL LIE IN INTER-SYSTEM COMMUNICATION, THE ALLOCATION OF LARGE TASKS TO MULTIPLE PROCESSORS AND PERFORMANCE. PERFORMANCE OF COMMERCIALLY AVAILABLE PACKAGES IS IMPROVING. THE COST-EFFECTIVENESS OF COMMERCIAL PACKAGES MAKES CUSTOM DEVELOPMENT OF EXPERT SYSTEM SHELLS DIFFICULT TO JUSTIFY FOR MOST APPLICATIONS. THIS PROPOSAL ADDRESSES THE PROBLEMS OF 1) MOVING REAL-TIME DATA TO/FROM MULTIPLE EXPERT SYSTEMS AND 2) GIVING THE APPLICATION PROGRAMMER FREEDOM TO EXPERIMENT EASILY WITH MESSAGE PRIORITIES AND TASK PRIORITIES/ALLOCATION. THE KEY TO THIS PROPOSED CONCEPT IS DESIGN OF A SET OF SOFTWARE TOOLS TO BE MADE AVAILABLE TO THE APPLICATION PROGRAMMER IN A MULTITUDE OF ENVIRONMENTS. THIS ALLOWS HIM TO DESIGN PRIORITIZED MESSAGES BETWEEN EXPERT SYSTEM AND REAL-TIME PROCESSES WITHOUT CONCERN FOR THE DETAILS OF IMPLEMENTATION. THIS ALSO FREES HIM TO CHOOSE THE BEST SHELL, PACKAGE OR CUSTOM SOFTWARE FOR EACH PIECE OF A PROBLEM.