Extending Flat Motion Planning to Non-flat Systems. Experiments on Aircraft Models Using Maple

Aircraft models may be considered as flat if one neglects some terms associated to aerodynamics. But some maneuvers may be hard to achieve with this flat approximation. Computational experiments in Maple show that in some cases a suitably designed feed-back allows to follow such trajectories, when applied to the non-flat model. We propose an iterated process to compute a more achievable trajectory, starting from the flat reference trajectory. More precisely, the unknown neglected terms in the flat model are iteratively re-evaluated using the values obtained at the previous step. This process may be interpreted as a new trajectory parametrization using an infinite number of derivatives, a property that may be called generalized flatness. We illustrate the pertinence of this approach in flight conditions of increasing difficulties, from power-off gliding flight to aerobatics.