Page 209 - Autonomous Mobile Robots
P. 209

194                                    Autonomous Mobile Robots

                                               n
                                                      m
                                with x ∈ U ⊂ R , u ∈ R , and m < n. Despite its simple formulation this
                                problem does not possess a simple solution, as can be inferred from the Theorem
                                of Brockett [2]. This theorem, yielding necessary conditions for smooth sta-
                                bilizability for general nonlinear systems, provides necessary and sufficient
                                conditions for feedback stabilizability of nonholonomic systems.


                                                                             n      m
                                Theorem 5.1 [2] Let ˙q = g(q)u be given, with q ∈ R , u ∈ R , g(q 0 )u 0 =
                                0, g(·) continuously differentiable in a neighborhood of q 0 . Assume, moreover,
                                that span{g(q)} is a nonsingular distribution of dimension m in a neighborhood
                                of q 0 . Then:


                                    1. There exists a continuously differentiable control law which makes
                                       (q 0 , u 0 ) asymptotically stable iff m ≥ n.
                                    2. There exists a continuously differentiable and dynamic feedback law
                                                 2
                                       which makes (q 0 , ξ 0 , u 0 ) asymptotically stable iff m ≥ n.

                                   The first part of Theorem 5.1 is due to Brockett [2], while the second
                                one to Pomet [10] (see also the work of Ryan [35] for a more general result in
                                the framework of nonsmooth stabilizability). We will not present the proof of
                                the above theorem, which can be found in the literature [2,10,36]; we simply
                                mention that the provided obstruction to stabilizability has a topological nature.
                                The essence of Theorem 5.1 is that the only interesting nonholonomic systems
                                are those for which the distribution g(q) drops dimension precisely at q 0 ,is
                                not continuously differentiable at q 0 , or is not defined at q 0 . In such cases
                                                                         1
                                we cannot infer anything about the existence of C (smooth), time invariant,
                                static or dynamic, asymptotically stabilizing control laws. Motivated by the
                                conclusions of Brockett’s Theorem we focus on:


                                     • State feedback control laws described by equations of the form

                                                              u = a(x)                  (5.2)


                                                      m
                                                n
                                      where α : R → R is a discontinuous function of its arguments.
                                     • State feedback, hybrid, control laws described by equations of
                                      the form

                                                      u = k(x, s d ),  s d = k d (x, ¯s d )  (5.3)

                                2  ξ denotes the state of the dynamic controller.




                                 © 2006 by Taylor & Francis Group, LLC



                                 FRANKL: “dk6033_c005” — 2006/3/31 — 16:42 — page 194 — #8
   204   205   206   207   208   209   210   211   212   213   214