Maximal feedback linearization and its internal dynamics with applications to mechanical systems on R4

International audienceFor systems that are not feedback linearizable, a natural question is: how to find the largest feedback linearizable subsystem and, if the partial linearization is not unique, what are the control-theoretic properties of various partial linearizations. In this paper, we will consider the problem of how to choose a partially linearizing output that renders the zero dynamics asymptotically stable and when such an output exists.We will state general results solving completely the problem for systems whose linearizability defect is one by identifying and describing two classes of systems. For the first class, all maximal partial linearizations lead to the same zero dynamics. For the second class, any asymptoticbehavior of the zero dynamics can be achieved by a suitable choice of a partially linearizing output. In the second part, we apply our results to mechanical systems with two-degrees-of-freedom and provide a detailed study of their partial linearizations.We illustrate the obtained results by examples of Acrobot (which belongs to the second class) and Pendubot (which belongs to the first class)