InNDIRECT methods of two-point boundary-value (TPBV)ntrajectory optimization, guidance, and control are generally builtnupon a calculus-of-variations foundation [1,2] using specified perfor-nmance indices written in the form of definite integrals of continuousnfunctions having continuous first and second derivatives betweenninitial and final times. Predictive shooting of trajectories is the toolnmost widely used to initialize the costate differential equationsnassociated with indirectly optimized TPBV solutions. In the interestnof fast, reliable initialization convergence, one seeks improvementsnthat provide 1) reductions of costate systemdifferential order, search-nspace dimensionality, andmultiple-costate interactions; 2) reductionsnof final-state sensitivities to small maladjustments of initial costatesnand unanticipated en route state changes; and 3) analytic propagationnof costates between time boundaries.
展开▼
