Page 162 - Mathematical Techniques of Fractional Order Systems
P. 162

150  Mathematical Techniques of Fractional Order Systems




                                     2            3
                                      Γ 11  Γ 12  Y
                                 Ω 5       Γ 22  T  5  $ 0           ð5:64bÞ
                                     4
                                                Z
            where

                           ~   ~  T          T  T  ~   ~  T   ~    ~
                   ϒ 11 5 A 0 P 1 PA 1 B 0 Y 0 1 Y B 1 Y 1 Y 1 τ m Γ 11 1 Q
                                 0           0  0
                             ~     ~  T   ~
                   ϒ 12 5 A τ 2 PY 1 T 1 τ m Γ 12
                           ~  T
                   ϒ 13 5 τ m PA Z
                             0
                                       T
                   ϒ 22 52 Λ 1 Q 2 T 2 T 1 τ m Γ 22
                            T
                   ϒ 23 5 τ m A Z
                            τ
                   ϒ 33 52 τ m Z
            and
                                                      ~
                                                             ~
                                                ~
                                                                  ~
                                      ~
                               ~
                     ~
                                            ~
                           ~
                 ~
                 Y 5 PY;   Q 5 PQ;   T 5 T P;   Γ 11 5 PΓ 11 ;  Γ 12 5 PΓ 12 :
            hold for a given scalars τ m and @τ m satisfying the Assumption (5.20),
            where   denotes the corresponding part of a symmetric matrix.
                                                                        ~
               The pseudo-state feedback controller gain matrix is given as K 0 5 Y 0 P.
            Proof 4: The proof of this Theorem is similar to the one given for the
            Theorem 2, the explanation here is omitted for brevity. After replacing the
            matrix A 0 by A 0 1 B u K 0 and following the same procedure as the proof of
            Theorem 3, i.e., pre- and postmultiply the obtained inequality by the follow-
                          21               21
                       2          3     2           3 T
                         P    0  0        P    0  0
            ing matrices  4  0  I  0  5  and  4  0  I  0  5  , respectively, and making
                          0   0  I         0   0  I
                                          ~    21
            some change of variables, where P 5 P  yields the inequality (5.64). It
            completes the proof.
               We can remark that the feedback stabilization problem based on the
            time-delay-dependent stability condition given by the inequality (5.64) is a
            nonconvex problem. This is due to the presence of decision variable Zalone
                            ~
                               T
            and in the product PA Z, such that the matrix A 0 is not necessarily nonsin-
                               0
            gular, which leads to a bilinear matrix inequality (BMI) structure. Then, the
            inequality (5.64) can not be solved for the given decision variables in the
            same time. Thanks to the Schur lemma, all diagonal components of a matrix
            must satisfy the inequality. Therefore, a two-step procedure is proposed to
            resolve this problem, as in Boukal et al. (2016a). Firstly, solving the first
                                                             ~
                                                                   ~
            component of the inequality (5.64) allows us to compute P and Y 0 . Then,
                                ~
                          ~
            after obtaining P and Y 0 , replacing them into the inequality (5.4) by their
            value leads to a feasible LMI.
   157   158   159   160   161   162   163   164   165   166   167