Page 433 - Biomimetics : Biologically Inspired Technologies
P. 433

Bar-Cohen : Biomimetics: Biologically Inspired Technologies  DK3163_c016 Final Proof page 419 21.9.2005 11:49pm




                    Biomimetic and Biologically Inspired Control                                419
                                                                   h
                                                                                   T
                                              ½              T                   d n Š , and k 0 .
                    which is parameterized by c 0 ¼ c 1  c 2      c n Š , d 0 ¼ d 1  d 2
                      With respect to the unknown system of (16.24)–(16.26), the feedforward controller Q(u)is
                    described by
                                                  dj (t)
                                                    1
                                                       ¼ Fj (t) þ gf d (t)                   (16:28)
                                                           1
                                                   dt
                                                  dj (t)
                                                    2
                                                       ¼ Fj (t) þ gu(t)                      (16:29)
                                                           2
                                                   dt
                                                  T
                                                                        T
                                                          T
                                          u ff (t) ¼ c j (t) þ d j (t) þ kf d (t) ¼ u j(t)   (16:30)
                                                   1
                                                            2
                    so as to generate the feedforward input u ff from the desired force signal f d (t). Here
                                                     T                           T
                                      u ¼ c T  d T  k  ,  j(t) ¼ j (t) T  j (t) T  f d (t)   (16:31)
                                                                      2
                                                               1
                    and the robot’s control input u(t) is then given by
                                                            T
                                            u(t) ¼ u ff þ u fb ¼ u (t)j(t) þ K(s)e(t)        (16:32)
                    Next, introducing a new state equation for the vector j (t) with respect to the control error input
                                                                 e
                    e(t) ¼ f d (t)   f(t)as

                                                  dj (t)
                                                    e
                                                       ¼ Fj (t) þ ge(t)                      (16:33)
                                                           e
                                                    dt
                    and defining
                                                                              T
                                         ~
                                                   T
                                        j j(t): ¼ j (t)   j (t) T  j (t) T  f d (t)   e(t)   (16:34)
                                                1      e      2
                    we can express the total control input u(t)as
                                                             T ~
                                                      u(t) ¼ u j(t)                          (16:35)
                                                             j
                                                             0
                    which is linearly parameterized by u 0 , see Muramatsu and Watanabe (2004) for the details of the
                    derivation.
                      Finally, to derive an adaptive rule, let us define

                                                               ~
                                                            T
                                                               j
                                                     ^ u u(t): ¼ u (t)j(t)                   (16:36)
                    by replacing u 0 in Equation (16.35) with u(t), and defining an error signal «(t)as
                                                                  ~
                                                                            T ~
                                                                 T
                                                        f
                                                                             j
                                                  u
                                       «(t): ¼ u(t)   ^ u(t) ¼ u 0   u(t)g j(t) ¼ c(t) j(t)  (16:37)
                                                                  j
                    where
                                                                                             (16:38)
                                                     c(t): ¼ u(t)   u 0
   428   429   430   431   432   433   434   435   436   437   438