Page 219 - Autonomous Mobile Robots
P. 219

204                                    Autonomous Mobile Robots

                                and it is discontinuous if one of the g i (z, zw) is such that g i (0, 0)  = 0. If the
                                system (5.16) is not discontinuous we can further transform it with a second σ
                                process. 12


                                5.3.3 The Issue of Asymptotic Stability
                                Theorems 5.3, 5.4, and 5.5 yield sufficient conditions for stabilizability of
                                discontinuous nonholonomic systems, while the σ process allows to map a
                                continuously differentiable system into a discontinuous one. To have a practic-
                                ally useful result, we have to show that asymptotic stability of the transformed
                                (discontinuous) system implies almost asymptotic stability of the original
                                (continuously differentiable) system. Moreover, to implement a discontinuous
                                control we must define it on the points of singularity.
                                   Consider a continuously differentiable nonholonomic system described
                                by equations of the form (5.1). Set x = col(x 1 , x 2 ) with x 1 ∈ R and
                                x 2 = col(x 21 , ... , x 2,n−1 ) ∈ R n−1  and define the σ process 13


                                                           ξ 1      x 1
                                                      ξ =     =                           (5.17)
                                                           ξ 2   σ(x 1 , x 2 )
                                where ξ 2 = col(ξ 21 , ... , ξ 2,n−1 ), σ(x 1 , x 2 ) = col(σ 1 (x 1 , x 2 ), ... , σ n−1 (x 1 , x 2 )),
                                               α i
                                and σ i (x 1 , x 2 ) = x /x β i  with α i ≥ 1 and β i ≥ 0, for all i = 1, ... , n − 1.
                                               2i  1
                                The application of the σ process (5.17) to the system (5.1) yields a new system
                                which is, in general, not defined for ξ 1 = 0. Suppose now that the transformed
                                system, with state ξ,is exponentially stabilized by a control law u = u(ξ),
                                that is, |ξ 1 (t)|≤ c 1 exp(−λ 1 t) and |ξ 2i (t)|≤ c 2i exp(−λ 2i t) for some positive
                                λ 1 , λ 2i , c 1 , and c 2i and for all i = 1, ... , n − 1. Then |x 1 (t)|≤ c 1 exp −λ 1 t  and
                                |x 2i (t)|≤ (c 1 c 2 ) 1/α i  exp(  −λ 1 β 1 +λ 2i t) for all i = 1, ... , n − 1. We conclude
                                                        α i
                                that exponential convergence to zero of the state ξ of the transformed system
                                implies exponential convergence to zero of the state x of the original system.

                                Remark 5.4   The previous conclusions also remain valid if the stabilizer is
                                dynamic. This fact is useful to design dynamic, output feedback, discontinuous
                                stabilizers for nonholonomic systems [46].
                                Remark 5.5   Asymptotic stability of the system with state ξ does not imply
                                asymptotic stability of the system with state x, as the inverse of the coordinates
                                transformation (5.17) does not map neighborhood of ξ = 0 into neighborhood
                                of x = 0, as illustrated in Figure 5.1. Therefore, exponential stability (in the
                                sense of Lyapunov) of the closed loop system with state ξ implies only almost
                                exponential stability of the closed loop system with state x.

                                12  Note that the composition of σ processes yields a σ process.
                                13  The coordinates transformation (5.17) defines a σ process only if   i β i ≥ 1.




                                 © 2006 by Taylor & Francis Group, LLC



                                FRANKL: “dk6033_c005” — 2006/3/31 — 16:42 — page 204 — #18
   214   215   216   217   218   219   220   221   222   223   224