BROADBAND TRACKING ALGORITHM DEVELOPMENT

PARAMETRIC APPROACHES TO TARGET TRACKING ARE STRAIGHTFOWARD AND USUALLY IMPLEMENTED VIA SOME VARIATION OF THE KALMAN FILTER. HOWEVER, A PARAMETRIC APPROACH USES SOME FORM OF A GRADIENT SEARCH ALGORITHM WHICH CAN BECOME LOCKED ONTO EXTRANEOUS OR AMBIGUOUS SOLUTIONS. THIS OCCURS BECAUSE THE OBJECTIVE FUNCTION IS MULTIMODAL. NOT ONLY CAN A PARAMETRIC APPROACH YIELD ERRONEOUS SOLUTIONS, BUT IT CANNOT DISCERN THAT IT HAS LOCKED ONTO A LOCAL RATHER THAN A GLOBAL SOLUTION. AN ALTERNATIVE TARGET TRACKING APPROACH IS PROPOSED FOR THIS EFFORT. THE SOLUTION PROPOSED IS BASED ON A NON-PARAMETRIC TECHNIQUE WHICH DOES NOT INVOKE A PRATICULAR STRUCTURE ON TARGET MOTION DYNAMICS. THIS APPROACH DISCRETIZES TARGET LOCATION TO A TWODIMENSIONAL GRID IN (X.Y), WHICH IS PROBABILISTICALLY UPDATED VIA BAYESIAN METHODS. THIS APPROACH EXAMINES THE LIKELIHOOD OF ALL POSSIBLE SOLUTIONS IN THE (X<Y) GRID, AND HENCE, IS ALWAYS SEEKING A GLOBAL SOLUTION. THE TARGET POSITION IS ESTIMATED EITHER AS AN MSE SOLUTION OR MAP SOLUTION.