PC-BASED COOPERATING SYSTEMS

ARTIFICIAL INTELLIGENCE RESEARCH HAS SHOWN THE IMPORTANCE OF COOPERATING EXPERT SYSTEMS ARCHITECTURES IN THE CONSTRUCTION OF INTELLIGENT PROGRAMS TO ASSIST WITH TASKS RANGING FROM DISTRIBUTED SITUATION ASSESSMENT TO INFERENCE OF PROTEIN STRUCTURE. THE MOST WIDELY ADOPTED IS THE BLACKBOARD ARCHITECTURE, WHICH HAS PROVEN HIGHLY EFFECTIVE IN A NUMBER OF SETTINGS. ONE SIGNIFICANT PROBLEM IS THE DEVELOPMENT OF SUITABLE TOOLS FOR THE EFFICIENT IMPLEMENTATION OF SYSTEMS UTILIZING SUCH ARCHITECTURES. A FEW TOOLS HAVE BEEN DEVELOPEDFOR MAINFRAME COMPUTERS. NONE AT ALL HAVE BEEN DEVELOPED FOR SMALL MACHINES SUCH AS THE IBM PC. HOWEVER, MANY REAL-WORLD INFORMATION ACQUISITION AND MANAGEMENT SYSTEMS (E.G., MILITARY SITUATION ASSESSMENT) ARE COMPOSED OF GROUPS OF SMALL MACHINES SUCH AS PCS NETWORKED WITH 32-BIT WORKSTATIONS AND MINICOMPUTER. THE SITUATION CREATES BOTH A DEMAND AND AN OPPORTUNITY TO DEVELOP TOOLS FOR IMPLEMENTING BLACKBOARD-BASED COOPERATING EXPERT SYSTEMS IN THE ENVIRONMENT OF NETWORK PCS AND WORKSTATION. THIS PROPOSAL ADDRESSES A FEASIBILITY STUDY FOR CONSTRUCTING SUCH TOOLS. THE TECHNICAL APPROACH IS BASED ON THE USE OF META-INTERPRETERS AND PARTIAL EVALUATION COMBINED WITH HIGH-SPEED INCREMENTAL PROLOG COMPILERS.