ADVANCED DEEP-STRIKE CRUISE MISSILE AUTONOMOUS TARGET-SCENE INTERPRETATION AND DECISION MAKING

ADVANCES IN AUTOMATED IMAGE UNDERSTANDING AND DECISION AIDS HAVE CREATED A VAST NEW POTENTIAL FOR SUPPORT OF STRATEGIC AND TACTICAL OPERATIONS. THE NON-MONOTONIC PROBABILIST (NMP), DEVELOPED BY MEMBERS OF THE PROPOSED PROJECT TEAM, IS A UNIQUE GENERIC INFERENCE FRAMEWORK WHICH IS DOMAIN INDEPENDENT AND CAN BE APPLIED TO A NUMBER OF PROBLEMS. NMP COMBINES THE SHAFER-DEMPSTER THEORY OF EVIDENTIAL REASONING WITH A FLEXIBLE ABILITY TO MAKE AND REVISE ASSUMPTIONS, TO EXAMINE THE DEGREE OF CONFLICT ASSOCIATED WITH THE CURRENT SET ASSUMPTIONS, AND TO BUILD IN CONFLICT RESOLUTION RULES THAT AUTOMATICALLY EXAMINE AND RESOLVE CONFLICTS WHEN THEY ARISE. IT IS A PARTICULARLY WELL SUITED FORMALISM FOR HANDLING CONFLICTING INFORMATION SOURCES SUCH AS THOSE THEY MAY BE ENCOUNTERED IN TARGETING PROBLEMS WHERE, FOR INSTANCE, CONFLICTS ARISE WHEN A PLANNED TARGET IS NO LONGER AVAILABLE. THE PROPOSED RESEARCH WILL APPLY THE CURRENT NMP ARCHITECTURE FOR EVIDENTIAL REASONING TO TARGETING PROBLEMS FOR RELOCATABLE OR MOVABLE TARGETS. THE ENSUING KNOWLEDGE BASE WILL PROVIDE A MODEL FOR ANALYZING THE COMPUTATIONAL STRUCTURE OF NMP WITH A VIEW TO INVESTIGATING APPROACHES TO DEALING WITH ITS INHERENT COMPUTATIONAL CONSTRAINTS. A PROBLEM APPROPRIATE FOR DEMONSTRATION WILL BE SELECTED AND A SMALL SCALE PROOF-OF-CONCEPT SYSTEM WILL BE DEVELOPED.