LEARNING SYSTEMS FOR ELECTRONIC COMBAT APPLICATIONS (LSECA)

ADVANCES IN ELECTRONIC COMBAT APPLICATIONS WILL SIGNIFICANTLY INCREASE THE ABILITY OF PILOTS TO OUTPERFORM A LARGE NUMBER OF ADVERSARIES IN ENVIRONMENTS WHERE THERE IS AN EXTREME INFORMATION PROCESSING BURDEN. THE NUMBER OF FACTORS PILOTS MUST CONSIDER IN MAKING ELECTRONIC COMBAT RELATED DECISIONS IS VERY HIGH AND OFTEN INVOLVE UNKNOWN RELATIONSHIPS THAT MUST BE LEARNED FROM EXPERIENCE. IN ADDITION, MUCH OF THE DATA IS UNRELIABLE OR MISSING, RESULTING IN A SUBSTANTIAL AMOUNT OF UNCERTAINTY. THIS PROPOSAL DISCUSSES A NEW FORM OF REASONING CALLED ABDUCTIVE REASONING AND RELATED INDUCTIVE LEARNING ALGORITHMS THAT PROVIDE A GENERAL MEANS TO ATTAIN SATISFACTORY SOLUTIONS TO PROBLEMS THAT CAN NOT BE RESOLVED USING CURRENT COMPUTER SCIENCE METHODS. THE LEARNING ALGORITHMS ARE BASED ON POLYNOMIAL NETWORK MODELING TECHNIQUES WHICH OFFER TREMENDOUS NEARTERM POTENTIAL IN AUTOMATICALLY SYNTHESIZING EFFECTIVE DECISION MODELS TO DEAL WITH THE UNCERTAINTIES ASSOCIATED WITH ELECTRONIC COMBAT APPLICATIONS, AND PROVIDE THE MEANS TO OBTAIN SATISFACTORY AND PRACTICAL REAL-TIME SOLUTIONS TO SUCH COMPLEX PROBLEMS.