Optimal control is recognized by the Airborne Wind Energy (AWE) community as a crucial tool for the development of the AWE industry. More specifically, the optimization of AWE systems for power generation is required to achieve the performance needed for their industrial viability. Models for AWE systems are highly nonlinear coupled systems. As a result, the optimization of power generation based on Newton-type techniques requires a very good initial guess. Such initial guess, however, is generally not available. To tackle this issue, this paper proposes a homotopy strategy based on the relaxation of the dynamic constraints of the optimization problem. The relaxed problem differs from the original one only by a single parameter, which is gradually modified to obtain the solution to the original problem.