Abstract : Optimal design of launch vehicles is a complex multidisciplinary design optimization (MDO) problem that has the distinctive feature of integrating a trajectory optimization which is very constrained, difficult to solve and highly coupled with all the other disciplines involved in the design process (e.g. propulsion, aerodynamics, structure, etc.). This PhD thesis is focused on the methods which allow to adequately integrate the optimal control law calculation into the design variable optimization process. A new method, called "Stage-Wise decomposition for Optimal Rocket Design" (SWORD), has been proposed. This method splits up the design process into the different flight phases and transforms the multistage launch vehicle optimization problem into the coordination of the optimizations of each of the stages, which are easier to solve. The SWORD method has been compared to the classical MDO method (Multi Discipline Feasible) in the case of the global optimization of a three-stage-to-orbit launch vehicle. The results show that the SWORD method allows to improve the efficiency of the optimization process concerning the feasible solution space search velocity and the quality of the optimum obtained in a limited calculation time. Moreover, an optimization strategy dedicated to the SWORD method has been developed. It allows to improve the convergence velocity of the method without requiring any a priori knowledge from the user with regard to the initialization and the search space.