A recently developed nonlinear programming stochastic projection of the gradient-type technique is applied to get and test novel procedures for tri-dimensional guidance and control of a satellite solid-fuel launch vehicle. Near optimal open-loop and low-frequency closed-loop solutions are obtained by parametrization of the control history and by considering the satisfaction of boundary constraints unless of specified random errors. Numerical simulations and tests show that the closed-loop solution can cope with the typical model uncertainties and environmental perturbations. A non standard situation of having a fourth stage with a free burning time is tested.