Engine control is a multi-objective task, in which an increasing system complexity and number of degrees of freedom are being used to meet higher and often conflicting performance requirements on several target quantities, which in most cases cannot be measured in production vehicles. The increasing number of degrees of freedom opens new possibilities but controlling such a complex system with strong cross couplings frequently leads to complicated trial and error procedures with a significant performance waste. Even though this is widely recognized and model based control is almost unanimously advocated as the solution of choice, the enormous legacy of decades of heuristic solutions makes the transition slow and uncertain. Against this background, this thesis suggests a systematic design and optimization approach which is proven both in simulation and experiments to allow significant performance improvements while reducing the design and calibration work and even simplifying the control and the measurement tasks. The price for it is higher but still affordable computational task and a truly interdisciplinary procedure, combining the understanding of combustion processes with modern control design approaches and new optimization techniques.