controller design for a class of cascade nonlinear systems using

Upload: dailyguyuk

Post on 03-Jun-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/12/2019 Controller Design for a Class of Cascade Nonlinear Systems Using

    1/8

    Controller Design for a class of Cascade Nonlinear Systems using

    Backstepping and Control Lyapunov Function Techniques

    Pan Feng Gu Rui

    (Control Science and Engineering Research Center, Southern Yangtze University,214122, P.R. China)

    Abstract:This paper studies controller design approaches for a class of cascade nonlinear systems achievinginput-to-state stability with respect to the bounded disturbance, which can be measured and used for control. A

    general form of using backstepping and control Lyapunov Function (CLF) techniques to design state feedback

    controllers for cascades is proposed. In this paper, two types of controllers (C1, C2) are discussed. Their

    properties are illustrated with simulation. Analysis shows that controller C1has a faster and better state

    response with disturbance input increasing than controller C2. On the other hand, controller C1is

    dramatically affected by the gain k, while controller C2hardly varies with the gain b varying.

    Keywords: Input-to-state stability, Cascade, State feedback control, Backstepping, Control Lyapunov

    Function .

    1. Introduction.Among studies on the stabilization of cascade systems, a particularly useful result is the

    input-to-state stability (ISS) condition of Sontag [10], which states that if subsystems of degree

    two are ISS, global asymptotical stability (GAS) respectively, then the general cascade is GAS.

    This result has been widely used for nonlinear design. In the literature several successful,

    constructive control methods were presented such as backstepping and forwarding (see e.g.

    Sepulchre et al[1], Praly et al[5], Arack et al[4], Mazenc.et al[3] ).

    In this paper, uniting ISS-CLF and backstepping techniques, we design two different state

    feedback controllers (C1, C2) for a class of cascade nonlinear systems achieving input-to-state

    stability with control and disturbance input. Analysis characters show that controller C1 has a

    faster and better state response with disturbance input increasing than controllerC2, but the state

    response of controller C1varies with the gain k. Adding disturbance restriction to the cascade

    systems, which vanishes at the origin, we finally render the cascade global asymptotically stable

    to the origin.

    This paper is organized as follows: in the next section we recall necessary definitions and

    review background results. In section 3 we design the stabilizing control laws for a class of

    cascade systems. An illustrative example is given in section 4. Some properties of two different

    controllers are compared by simulations. Finally, we get some conclusions in section 5.

    2. Preliminaries.

    A function RRr : is said to be of class K if it is continuous, strictly increasing and

    is zero at zero. It is of class K if, in addition, it is unbounded.

    A function

    RRR : is said to be of classKL if, for each fixed s , the function

    ),( t decreases to zero at infinity.

    stands for the Euclidean norm.

    o stands for composition. )( 2121 rrrr =o

    Function RRh : and RRf : , the Lie derivative of h alongf , written as hLf ,

    http://www.paper.edu.cn

  • 8/12/2019 Controller Design for a Class of Cascade Nonlinear Systems Using

    2/8

    1

    is defined by )()( xfx

    hxhLf

    = .

    A general nonlinear system of the form

    dxgxfx )()( +=& (1)

    where f is a local Lipschitz function and d is a locally essentially bounded disturbance

    input.

    Definition1: System (1) is said to be input-to-state stable if for any bounded disturbance d

    and initial state )0(x , the solution )(tx satisfies the following estimate:

    )(),)0(()( drtxtx + (2)

    where is of classKL , r is of classK.

    Theorem 1: System (1) is input-to-state stable if and only if there exists an ISS-CLF

    RRxV n :)( , a smooth positive definite radially unbounded function, holds for all 0x :

    dxdxdVLVLgf

    ,)()()( ++ (3)

    3. Controller design of state feedback control

    Consider a cascade nonlinear system of the form

    +=++=

    +=++=

    dxxguxxfuxxdxxgxxfx

    dxgxxfxxdxgxfx

    ),(),,(),(),(),(

    )(),()()()(

    2122122122122122

    1121121111111

    &

    &

    )5(

    )4(

    This system is represented by the block-diagram, which shows a feedback loop. We need to find

    a control law ),( 21 xxuu= via backstepping that ensures the input-to-state stability for the

    closed-loop cascade systems.Proposition 1: Consider the system of form:

    uxdxgxfx )()()( ++=& (6)

    with the following assumptions.

    1) System (6) is input-to-state stable.

    2) There is a smooth control law ),( dxRu= satisfying 0)0,0( =R .

    Under these conditions, the augmented system (4),(5) is input-to-state stable with a smooth

    control law ),,( 21 dxxRu=

    Proof: We define the following control law ),( 111 dxRu = paralleling the correspondingSontag-type control law )(xRu= [4]

    0

    0

    0

    )(

    ),(1

    1

    1

    4

    1

    2

    11

    1

    1

    1

    1

    =

    ++=

    VL

    VL

    VL

    VLww

    dxR

    (7)

    Where dVLVLw gf )( 11 11 += , the disturbance can be measured and used for control.

    The above control law is smooth on { }0\nR . By theorem 1, we know that systems (4) is ISS,

    then it admits a smooth ISS-CLF )(11

    xV so that we have

    dxdxdxRVLdVLVL gf ,).()(),()()( 1111111 111 +++ .

    http://www.paper.edu.cn

  • 8/12/2019 Controller Design for a Class of Cascade Nonlinear Systems Using

    3/8

    2

    First, we design controller C1: ),,( 212 dxxRu= .

    From above, we know the system (5) can be stabilized by a smooth state feedback control

    law ),( 112 dxRx = with 0)0,0( =R . Rewrite the cascade as

    [ ] [ ]

    ++=

    +++=

    uxxdxxgxxfx

    xRxxdxRxdxgxfx x

    ),(),(),(

    )()(),()()()(

    2122122122

    1121111111111

    &

    &

    )9(

    )8(

    The change of variable ),( 112 dxRxz =

    Results in the system

    [ ]

    =

    +++=

    ),(

    )(),()()()(

    112

    1111111111

    dxRxz

    zxdxRxdxgxfx x&&&

    &

    The derivative ),( 11 dxR& can be expressed by

    211111 )()(),( 111 xRLdRLRLdxR gf ++=& (10)

    Taking ),( 112 dxRxv = &

    reduce the system to the cascade[ ]

    =

    +++=

    vz

    zxdxRxdxgxfx x

    &

    & )(),()()()( 1111111111

    Now construct the positive composite function

    2

    112122

    1)(),( zxVxxV += (11)

    The derivative2V along the solution of the system satisfies

    [ ]

    zvzVLdx

    zvzVLRVLdVLVLxxV gf

    +++

    ++++=

    )()()(

    )()()(),(

    11

    11111212

    1

    1111

    &

    choosing 0,11 >= kkzVLv

    yields

    )()(),((

    )()),(()()()(

    2

    1

    112

    1

    2

    1121

    2

    12

    dx

    xd

    dxRxk

    x

    ddxRxkxkzdxV

    +

    +

    ++=+&

    (12)

    where K .

    From above, we know ),( 212 xxV is an ISS-CLF for cascade system . Substituting zv,

    and 1R& , we obtain the state feedback control law of system

    2111121

    1

    2212 )()()[,(),,( 1111 xRLdRLRLVLxxdxxRu gf +++==

    ]),(),())(( 212212112 dxxgxxfxRxk (13)

    Thus, the control law ),,( 2122 dxxRu = achieves ISS of .

    Next, we design the second controller C2: ),,( 2122 dxxRu = .

    We define the positive composite function

    211211212 )),((

    21)(),( dxRxxVxxV += (14)

    The derivative2V along the solution of the system satisfies:

    http://www.paper.edu.cn

  • 8/12/2019 Controller Design for a Class of Cascade Nonlinear Systems Using

    4/8

    3

    ))(()()( 121221112 111 RxRxxVLdVLVLV gf&&& +++=

    { )()()()( 222121111 111 udgfRxRVLdVLVL gf +++++=

    }12111 1111 ])()([ VLxRLdRLRL gf +++ (15)

    We now choose the control u such that the terms except containing disturbance terms are

    controlled inside the bracket (15). In order for a controller to avoid less nonlinear cancellation

    than controller C1, it takes the form:

    ])([),,( 212111

    2212 111vfVLxRLRLdxxRu f ++==

    (16)

    yields

    vRxdRLRxdgRxdxV g )())(()()()( 1211221212 1 +++ & (17)

    Taking 0),()()( 122

    112

    2

    212 1>= bRxbRLRxgRxv g

    yields

    ])()[()()( 2122

    221212 dgRxgRxdxV + &

    2

    12121

    2

    1

    2

    12 )(]))(()[( 11 RxbdRLxRRLRx gg (18)

    By using Youngs inequality [9], we have

    dgRxdgRx 212222

    12 )(4

    1)( +

    dRLxRdRLRx gg ))((4

    1)( 121

    22

    1

    2

    12 11+

    Applying the inequalities above to (18) we get

    [ ]

    )(~~)(~

    )(

    ~

    )(~)()(2

    1)()()(

    2

    1

    12

    1

    2

    121

    22

    1212

    dx

    xd

    Rxb

    x

    dRxbxdRxbdxV

    +

    +

    ++=++&

    (19)

    Where K~,~ . Therefore ),( 212 xxV is an ISS-CLF for the system . This implies that the

    control law ),,( 212 dxxRu= input-to-state stabilizes cascade system . Thus, the controller C2

    should be of the form

    2

    1221211

    1

    2212 )()([),,( 111 gRxfVLxRLRLdxxRu f +==

    )]()( 122

    112 1RxbRLRx g (20)

    Remark 3.1: Only if the external disturbance is bounded, the controllers C1, C2can achieve ISS.

    Remark 3.2: From the solution )(tx estimation (2) , 0 as t . Since for 0tt we

    have )()(

    drtx . If there exists a class K function such that for all 0x , the inequality

    (3) holds:

    0)()(

  • 8/12/2019 Controller Design for a Class of Cascade Nonlinear Systems Using

    5/8

    4

    Remark 3.3: Although 2C is significantly more complex than 1C , the backstepping approach

    has its merit such that repeated application of above proposition, we can design ISS control laws

    for extended cascade systems of the form.

    ++=

    ++=

    ++=

    uxxdxxgxxfx

    xxxdxxgxxfx

    xxdxgxfx

    nnnnnnn ),,(),,(),,(

    ),(),(),(

    )()()(

    111

    32122122122

    21111111

    LLL&

    M

    M

    &

    &

    4. Example

    For a clearer illustration of some properties of two different controllers, we consider the

    following example of the cascade nonlinear system of the form:

    )23(

    )22(:1

    2

    21212

    2

    12

    211

    2

    11

    ++=

    ++=

    uxxdxxxxx

    xxdxxx

    &

    &

    Where d is external disturbance input.

    A Lyapunov function for (22) is 21112

    1)( xxV = .

    Using the equality (7), we obtain a control law

    =

    ++++

    ==

    0,0

    0,)()()(

    ),(

    1

    1

    1

    4

    1

    2

    1111

    111

    1

    1

    1

    11111

    VL

    VLVL

    VLdVLVLdVLVL

    dxRugfgf

    =

    ++++=

    0,0

    0,)()][(

    1

    1

    2

    1

    2

    11

    x

    xxdxdx

    The derivative ),( 11 dxR& can be expressed:

    ])(

    21[

    2

    1

    2

    1

    1

    1

    1

    xdx

    dx

    x

    u

    ++

    ++=

    (C1) From control law of (13), we obtain controller:

    2

    1211

    2

    2

    1

    2

    1

    1

    2

    21

    212 ))()(

    21(1),,(11

    xxxdxxxdx

    dxxx

    dxxRuC

    ++++

    ++==

    +++++ dxxxxxdxdxxk 2122

    1

    2

    1

    2

    112 ])()([

    (C2) From control law of (20), we obtain the controller

    2

    2

    1

    2

    121

    2

    12

    1

    2

    1

    1

    2

    21

    212 ))()(

    21(

    1),,(

    2xxxxxx

    xdx

    dx

    xxdxxRuC +

    ++

    ++==

    ++

    ++++

    +++++

    2

    2

    1

    2

    1

    111

    2

    21

    2

    1

    2

    112

    )(

    )2(])()([

    xdx

    dxxxxxbxdxdxx

    http://www.paper.edu.cn

  • 8/12/2019 Controller Design for a Class of Cascade Nonlinear Systems Using

    6/8

    5

    The dynamic controllers21

    , CC uu provide stabilization for the cascade nonlinear system 1 .

    Performances of the two controllers are compared by simulations. The transient responses of the

    designed closed-loop system are shown in figure 1

    (a) (b)

    Figure1. State response of controlled system without disturbance

    Remark 4.1: The response is computed with d=0 and Tx ]1,1[)0( = . Figure 1(a) represents the C1

    controller with )20,10,5(),,( 210 =kkk and Figure 1(b) represents C2 controller

    with )20,10,5(),,( 210 =bbb . The state response of the system driven by C1control law 1Cu has the

    same performance as the state response of the system driven by C2control law2C

    u achieving

    ISS. Note that the state response of controller C1results in the same performance for any positive

    values of the gain ),,( 210 kkk . So does controller C2.

    (a) (b)

    Figure2. State response of controlled system with the same disturbance

    Remark 4.2: All the initial conditions are the same as in figure 1. Figure 2 (a) represents the

    state response of controller C1that shows a faster and better damped response of the state with

    the gain k increasing which the equilibrium point of the perturbed system 1 is not the

    origin 0=x because of nonvanishing perturbation. Figure 2 (b) represents the state response of

    controller C2 that has similar trajectories to figure 1 (b) besides the equilibrium 02 x of the

    state 2x because of nonvanishing perturbation.

    http://www.paper.edu.cn

  • 8/12/2019 Controller Design for a Class of Cascade Nonlinear Systems Using

    7/8

    6

    (a) d=2 (b) d=5

    (c) d=10

    Figure3. State responses of controlled system with different disturbances

    Remark 4.3:The response is computed with initial conditions Tx ]1,1[)0( = , 10=k and 1=b .

    Solid curves are the state response of the system driven by C1control law1C

    u and dotted curves

    are the state response of the system driven by C2control law2C

    u . Figure 3 shows that the state

    1x s response of the two controllers have the same performance, but the state 2x s response of

    controller C1have the faster and better damped performance than controller C2 with the external

    disturbance increasing.

    Figure4. State response of controlled system with the same restricted

    )1(2

    2

    1

    xxd += disturbance

    Remark 4.4:Figure 4 shows that if the disturbance is restricted such as )1( 22

    1 xxd += , the two

    http://www.paper.edu.cn

  • 8/12/2019 Controller Design for a Class of Cascade Nonlinear Systems Using

    8/8

    7

    controllers have the same state response converging to the origin which the

    disturbance )1( 22

    1 xxd += vanishes at the origin.

    5. Conclusion

    In this paper we design two types of controllers of the cascade nonlinear system.

    Performances of these controllers are compared by simulations. We find that each of thecontrollers has its merits respectively. Controller C1shows a faster and better damped response

    of the state with the gain k and disturbance increasing, while controller C2 shows a slower

    response of the state than controller C1with the disturbance increasing and it hardly vary with

    the gain k varying. For some choices of the parameters such as limited disturbance which

    vanishes at the origin, the two controllers result in the same performance of the state response

    converging to the origin.

    REFERENCES

    [1] R. Sepulchre, M. Jankovic, and P.V. KoKotovic, Constructive Nonlinear Control. New York: Spinger-Verlag,1997

    [2] M. Jankovic and P.V. Kokotovic, Global stabilization of an enlarged class of cascade nonlinear systems

    [3] F. Mazeac, R. Secpulchre, and M. Jankovic, Lyapunov Functions for stable cascades and applications to Global stabilization.

    IEEE Trans. Automat. Contr.,Vol.44, pp1795-1800, sep. 1999.

    [4] M. Arack, A.R. Teel, P.V. Kokotovic, Robust Nonlinear control of feedforward systems with unmodeled dynamics.

    Automatica, Vol 37,pp 256-272,2001

    [5] L. Praly, R. Ortega, and G. Kaliora, Stabilization of Nonlinear system via forwarding mod{ VLg }. IEEE Trans. Automat.

    Contr, Vol.46. pp1461-1466. 2001.

    [6] E.D.Sontag. A universal construction of Artsteins theorem on nonlinear stabilization. Systems Control Lett. 13

    pp117-123.1989

    [7] A.Isidori, Nonlinear Control System II, London, 1999.[8] H.K.Khalil. Nonlinear Systems. Prentice Hall, Englewood Cliffs, NJ, Second edition, 1996.

    [9] G.Hardy, J.E. Littlewood, and G. Ploga Inequalities, 2nd.ed. Cambridge, UK, Cambridge Univ Press, 1989.

    [10] E.D. Sontag. Smooth stabilization implies coprime factorization. IEE. Trans. Automat. Contr. Vol 34. pp435-443. 1989.

    http://www.paper.edu.cn