USE OF POLYNOMIAL NETWORKS TO IMPROVE ON CONTROL EFFICIENCY OF MANEUVERABLE PROJECTILE

EXECUTION OF OPTIMAL GUIDANCE AND CONTROL ALGORITHMS IN REAL TIME WITH A MICROPROCESSOR IS POSSIBLE USING POLYNOMIAL NETWORKS TO STORE, IN COMPACT AND ALMOST INSTANTLY RETRIEVABLE FORM, INFORMATION ABOUT A LARGE NUMBER OF PRE-COMPUTED OPTIMUM TWO-POINT BOUNDARY-VALUE (TPBV)SOLUTIONS. THESE SOLUTIONS MAY BE OBTAINED OFF-LINE USING THE CALCULUS OF VARIATIONS, ITERATIVE SEARCH, DYNAMIC PROGRAMMING, AND EQUIVALENT METHODS, ALL LEADING TO POWERFUL BUT (IN THEIR USUAL FORMS OF IMPLEMENTATION) COMPLEX GUIDANCE AND CONTROL ALGORITHMS. SOLUTION OF A TPBV PROBLEM REQUIRES INITIALIZATION OF A FAMILY OF NONLINEAR DIFFERENTIAL EQUATIONS. HISTORICALLY, THESE INITIALIZATIONS HAVE NECESSITATED ITERATIVE COMPUTATIONS BECAUSE THE DIFFERENTIAL EQUATIONS ARE NOT, IN GENERAL, ANALYTICALLY INTEGRABLE. IT HAS RECENTLY BEEN DEMONSTRATED THAT THE APPROPRIATE INITIALIZATIONS CAN BE PERFORMED BY POLYNOMIAL NETWORKS THAT ARE FITTED, OFF-LINE, TO OPTIMUM TRAJECTORY SIMULATION DATA. IT IS PROPOSED THAT POLYNOMIAL NETWORKS BE APPLIED TO GUIDANCE AND CONTROL OF A MANEUVERABLE PROJECTILE THAT CAN ACHIEVE SIGNIFICANT PERFORMANCE BENEFITS USING A COMPLEX GUIDANCE AND CONTROL ALGORITHM IMPLEMENTED IN A SIMPLE AND ECONOMICAL FORM.