ESA GNC Conference Papers Repository

In this work, it is shown how a computationally-intelligent mechanism can produce robust controllers according to a multi-objective scheme employing genetic algorithms (to generate, reproduce, mutate and select individuals) and fuzzy systems (to score those individuals according to the specifications on stability, performance, interpolation, etc.). At the same time, the observer-based realization of each controller is computed according an on-board model composed of the physical variables of the plant plus other variables of interest, such as disturbances or faults; such technique allows a given controller to behave also as an observer (useful for fault tolerance purposes), without any change of the controller dynamics. A launch vehicle model is used to demonstrate these concepts, where non-linear digital simulations include sensor noise, sensor faults and external disturbances.