ARTIFICIAL INTELLIGENCE FOR COMMAND AND CONTROL

THE MODERN BATTFIELD CALLS FOR HIGH-STAKES DECISIONS AND JUDGMENTS IN AN INCREASINGLY COMPLEX ENVIRONMENT CHARACTERIZED BY HIGH VOLUMES OF UNCERTAIN, INCOMPLETE, AND OFTEN INCONSISTENT INFORMATION. IN THE FUTURE, INTELLIGENT COMPUTERIZED SYSTEMS CAPABLE OF ADAPTIVE LEARNING WILL BECOME NECESSARY AS AIDS TO HUMAN DECISION MAKERS. DSC HYPOTHESIZES THAT SIGNIFICANT ADVANCES IN THE CAPABILITY FOR TRUE ADAPTIVE LEARNING CAN BE ACHIEVED BY COMBINING A GOAL-BASED, HIERARCHIAL KNOWLEDGE REPRESENTATION WITH A STATE-OF-THE-ART CAPABILITY FOR REPRESENTING AND MANIPULATING UNCERTAINTY. THIS COMBINATION WILL MAKE POS-SIBLE A SYNERGISTIC LEAP IN THE FLEXIBILITY AND ADAPTABILITY OF INTEL-LIGENT BATTLEFIELD EXPERT SYSTEMS. FOUR TASKS ARE PROPOSED FOR PHASE I: (1) DEVELOP A THEORY OF HIERARCHICAL, GOAL-STRUCTURED KNOWLEDGE REPRESENTATION; (2) DEVELOP THEORIES FOR REPRESENTING AND MANIPULATING UNCERTAINTY; (3) SYNTHESIZE THEORIES OF KNOWLEDGE REPRESENTATION AND UNCERTAINTY INTO DESIGNS FOR EXPERT SYSTEM ARCHITECTURE; AND (4) IMPLEMENT SELECTED IDEAS IN A SMALL-SCALE PROTOTYPE SYSTEM.