PASSIVE TMA USING ITERATED KALMAN FILTERS AND GAUSSIAN SUMS

DHWA PROPOSES TO DEVELOP ALGORITHMS FOR LONG-RANGE, PASSIVE TARGET MOTION ANALYSIS (TMA). THE ALGORITHMS WILL USE EXTENDED FILTERS SIMILAR TO THOSE FOUND IN CURRENT ALGORITHMS (JASA, MPKAST, AND MTST), BUT WILL UNIFY AND EXTEND THESE TECHNIQUES IN IMPORTANT WAYS. WE HOPE TO ELIMINATE MOST OF THE PROBLEMS ASSOCIATED WITH CURRENT METHODS. THE DIFFERENCES AMONG CURRENT ALGORITHMS, AND THEIR LIMITATIONS, ARE USUALLY TRACEABLE TO THE COORDINATE SYSTEM AND THE POINTS AT WHICH LINEARIZATIONS ARE PERFORMED. THREE PRINCIPLES EMERGE: A.LINEARIZATIONS ARE UNAVOIDABLE; B.LINEARIZATION POINTS SHOULD BE CONSISTENT--SIMULTANEOUS LINEARIZATIONS SHOULD BE PERFORMED ABOUT THE SAME POINT, SEQUENCES OF LINEARIZATIONS SHOULD BE PERFORMED ABOUT POINTS WHICH FORM A TRACK CONSISTENT WITH THE MOTION MODEL AND THE DATA; AND C.THE FINAL SOLUTION SHOULD BE INDEPENDENT OF THE COORDINATE SYSTEM USED TO EXPRESS IT. WE WILL USE AN ITERATIVE PROCEDURE WHICH PRODUCES THE LIKELIHOOD TARGET TRACK INDEPENDENT OF THE COORDINATE SYSTEM. AN INFORMATION MANAGER WILL COMBINE THE RESULTS OF SEVERAL TRACKERS (OPERATING UNDER DIFFERENT HYPOTHESES) TO PRODUCE A COMPOSITE, NON-GAUSSIAN TMA SOLUTION. THE SYSTEM WILL ACCOUNT FOR CRITICAL INFORMATION WHICH GOES UNPROCESSED IN CURRENT ALGORITHMS. EXAMPLES ARE: ACOUSTIC PREDICTIONS; PERIODS OF GAIN, LOSS, OR HOLDING; AND KNOWN OPERATING CHARACTERISTICS.