A number of algorithms have been proposed to solve constrained optimization problem using evolutionary computation methods. However, most of them have not been useful when addressing heavily constrained optimization problems in terms of computational efficiency and solution accuracy. These algorithms have mostly been applied to evolution of single group. In this paper, we propose a new co-evolutionary algorithm based on the augmented Lagrangian methods. The proposed method is based on the evolution of two groups with opposite objectives. One group try to minimize the object parameters, the other try to maximize the Lagrange multipliers. Then the optimal solution is saddle point. The proposed method is applied to the autopilot design of a flexible launch vehicle problem.