This work presents a methodology for shape optimization of geometrically nonlinear structures. The main purpose is to avoid the stability problems generated by optimization based on linear behavior. The methodology was implemented for two-dimensional problems, and several structures were successfully optimized. Using geometrical modeling concepts, the shape of the structure is defined by its boundary curves. Therefore, parametric representation and curve definition by a set of key points are discussed in detail. Due to its flexibility in shape definition, particular attention is given to interpolation using B- splines. The optimization problem is defined based on the geometrical model and the design variables are the positions of key points. Design variable linking can be applied to enforce symmetry.The structure it is analyzed using plane isoparametric elements. Thus, is necessary to perform the discretization of the structure in a set of finite elements. Different algorithms were implemented to generate structured or unstructured finite element meshes. The standard Newton-Raphson method is applied to find the equilibrium configuration, and different methods can be used to evaluate critical points. Due to the convergence problems presented by direct methods, a new semi-direct method was developed. The numerical results show the suitability of the finite elements and numerical methods implemented in the present work.The mathematical programming algorithms used in this work require the evaluation of design sensitivities in order to compute the search direction of the optimization process.Using basic sensitivity equations, which are independent from the particular element, analytical expressions were developed for the sensitivity computation of isoparametric elements formulated according to the Total Lagrangian approach. Applying the analytical method for more complex elements is very cumbersome and error prone. On the other hand, the finite difference method is simple and generic, but its computational cost is prohibitive. The semi-analytical method preserves the advantages of the use of finite differences and has a low computational cost, but presents severe accuracy problems. Hence, a method based on the exact differentiation of the rigid body motions was developed in this work to improve the accuracy of the semi- analytical sensitivities of geometrically nonlinear structures. The numerical examples show that this method eliminates the abnormal errors presented by the semi- analytical sensitivities.