In mechatronics and robotics one often encounters highly complex nonlinear systems, whose dynamics are described by a system of nonlinear ordinary differential equations. In simulation, one can use very sophisticated models, whereas controllers are typically designed on strongly simplified models. Quite often, this results in linear controllers which are unsuitable for the original nonlinear system. We suggest an approach allowing nonlinear controller design even for very complicated models. Our approach is based on an alternative differentiation method called algorithmic or automatic differentiation. With this computation method we can circumvent problems that may arise in symbolic computation.