Intelligent Real-time Problem Solvin: Issues and Examples
Paul R. Cohen, Adele E. Howe, David Montgomery Hart · 1990
THIS REPORT PRESENTS OUR WORK ON REAL TIME PROBLEM SOLVING (IRTPS). THE TOPIC IS FUNDAMENTALLY CHALLENGING IN THE SENSE THAT IT PROBABLY CANNOT BE COMPLETELY ADDRESSED WITHIN THE ESTABLISHED KNOWLEDGE-BASED AND LOGISTIC PARADIGMS, BUT WILL REQUIRE METHODOLOGOGICAL, THEORETICAL, AND TECHNICAL DEVELOPMENTS. ACCORDINGLY, THIS REPORT LOOKS AT IRTPS FROM ALL THESE PERS- PECTIVES. BECAUSE READERS WILL HAVE DIFFERENT INTERESTS, EACH SECTION OF THE REPORT IS INDEPENDENT OF ALL SECTIONS EXCEPT THIS INTRODUCTION. SECTION 2 OFFERS A DEFINITION OF IRTPS. SECTION 3 DESCRIBES OUR REAL-TIME TESTBED, INCLUDING ITS CURRENT STATUS AND PORTABILITY, AND OUR TIMETABLE FOR MAKING IT GENERALLY AVAILABLE. SECTION 4 DISCUSSES THE ARCHITECTURE WE HAVE DEVEL- OPED FOR REAL-TIME AGENTS AND OUR NEAR-TERM RESEARCH GOALS. SECTION 5 IS DEVOTED TO METHODOLOGICAL ISSUES, SPECIFICALLY, HOW CHARACTERISTICS OF EN- VIRONMENTS CONSTRAIN THE DESIGN OF AGENTS (INCLUDING AN ASSESSMENT OF THE PROS AND CONS OF SIMULATED ENVIRONMENTS), HOW TO EVALUATE IRTPS SYSTEMS, AND THE NEED FOR ANALYTIC MODELS OF AGENT ARCHITECTURES. THE TASK ENVIRONMENT FOR MUCH OF OUR RESEARCH IS A SIMULATION OF FOREST FIRES. THE TASK IS TO CONTROL SIMULATED FIRES BY DEPLOYING SIMULATED AGENTS INCLUDING SMART BULLDOZERS, FUEL CARRIERS, AND AIRPLANES. (SMART AGENTS HAVE THE SIMULATED PHYSICAL ABILITIES OF, SAY, BULLDOZERS, AND SOME OF THE SIMULATED MENTAL ABILITIES OF THEIR HUMAN OPERATORS.) THIS IS A REALTIME