1872

ieee transactions on ultrasonics, ferroelectrics, and frequency control, vol. 53, no. 10, october 2006

Trajectory Tracking of Piezoelectric Positioning Stages Using a Dynamic Sliding-Mode Control

Hsin-Jang Shieh, Member, IEEE, and Po-Kai Huang
AbstractTrajectory tracking performance of a piezoelectric positioning stage almost depends on whether the tracking controller can eectively compensate the inherent hysteresis phenomenon. In this paper, a dynamic slidingmode control (DSMC) with backstepping is proposed for the trajectory tracking of the piezoelectric positioning stage, which is suitable for a component of scanning microscopes. An equivalent model developed from a linear motion dynamics with addition of the hysteresis nonlinearity and strain-dependent function rst is proposed to approximately represent the dynamics of motion of a onedimensional piezoelectric positioning stage. Then, based on the equivalent model, the DSMC with an asymptotical sliding surface is proposed for the trajectory tracking control of the piezoelectric positioning stage. Moreover, the analysis of stability can be completed by mathematics, and the convergence rate of the tracking error can be governed by the choice of the control parameter values. Using the DSMC to trajectory tracking control, the piezoelectric positioning stage becomes more suitable for practical applications, especially with the need of various trajectories tracking in microscopy. To validate the proposed control scheme, a computer-based controller and a piezoelectric positioning stage with a capacitive displacement sensor are implemented. Experimental results illustrate the feasibility of the proposed controller for trajectory tracking applications.

I. Introduction ecently, due to the requirement of the micrometer or nanometer resolution in displacement, high stiness, and fast frequency response, piezoelectric actuators often are used in many precision positioning applications, such as scanning tunneling microscopy [1], near-eld optical scanning microscopy [2], [3], and vibration control [4]. However, because the materials of piezoelectric actuators are ferroelectric, they fundamentally exhibit hysteresis behavior in response to an applied electric eld [5], [6]. Unfortunately, this behavior usually leads to problems with severe inaccuracy [4][6] and deteriorated tracking performance [7][10] when the piezoelectric are operated in an open-loop mode. Therefore, there have been many modeling techniques in hysteresis behavior [11][17] by mathematical functions or equations, such as the Preisach function [11][14] and the nonlinear, piecewise circuit description [15][17]. However, these modeling techniques were either frequency dependent and complicated or nonsystematic. From the viewpoint of system control engineer-

Manuscript received April 25, 2005; accepted May 12, 2006. The authors are with the Department of Electrical Engineering, National Dong Hwa University, Shoufong, Hualien 97401, Taiwan (e-mail: hjshieh@mail.ndhu.edu.tw). Digital Object Identier 10.1109/TUFFC.2006.119

ing, these cases lead to a large diculty in model-based controller design that is often used for system control engineers. Modeling techniques on piezoelectric devices for positioning control or manipulation have been presented in [8], [9], [18][22]. In [8], the hysteresis behavior described by the Preisach model was compensated by the proportionalintegral-derivative (PID) feedback controller incorporated with the feedforward loop. Observing their experimental results of the sine-wave tracking responses of the piezoelectric actuator, the hysteresis eect seemly was compensated. However, the controller design was independent of the hysteresis dynamics described by the used Preisach model. Moreover, due to the complicated computation and iterative integrals of the Preisach function, this method made the analysis of transient response and system stability more dicult. In [9], [18], the model consisting of several elasto-slide elements with the massless blocks and massless springs subjected to Coulomb friction were used to approximately describe the dynamics of motion of piezoactuators, and the positioning controllers were composed of the PID, feedback linearization, and repetitive control schemes. Although the dynamics of motion of piezoelectric actuators with hysteresis behavior could be approximately represented, the system parameters used in this model such as the number of the massless blocks and springs, break force, and stinessneeded to be accurately identied by experienced experimental tests. In [19], the hysteresis behavior was modeled by a set of hysteresis operators that included a gain and an input-dependent lag. The tracking controller was constructed by a traditional proportional-integral (PI) controller. Unfortunately, there were many limitations and assumptions in this modeling approach, such as the rate of change in control input voltages, the decision of input-dependent lag, and the restrictions on adoption of some parameters. In addition, the design of the traditional PI-type tracking controller was independent of the dynamics of motion of the piezoelectric actuator; that is, the modeling of the piezoelectric actuator in [19] seemly was unnecessary for controller design. In [20][22], the mechanical model consisting of the massspring-damper systems was investigated. In these models, the dierential equations related to the mechanical motion dynamics with the specied nonlinear terms were used to describe the motion dynamics of a piezoelectric translation mechanism with hysteresis behavior. Although these dynamic models could approximately represent the motion dynamics of a piezoelectric translation device or mechanism, the information about how to design a highperformance positioning controller was not provided.

08853010/$20.00 c 2006 IEEE

shieh and huang: trajectory tracking using dynamic sliding-mode control

1873

To cope with the drawbacks mentioned above, an equivalent model for describing the dynamics of motion of a piezoelectric positioning stage must be developed, then a model-based trajectory tracking controller that can provide an improved transient performance is addressed. A nonlinear dierential equation for one-dimensional motion of piezo-actuators in [22] is used and modied to represent the dynamics of motion of the piezoelectric positioning stage studied in this paper. Moreover, a thirdorder state equation in which all the states are available is proposed. To demonstrate the validity of the proposed dynamic model, experimental results of the frequencydependent hysteresis responses of both the mathematical model and actual system are illustrated. After this validation, a model-based dynamic sliding-mode control (DSMC) technique for trajectory tracking of the piezoelectric positioning stage is proposed. The reasons for using the DSMC to tracking controller design can be summarized as follows. The advantages of the conventional SMC [23], [24], such as improvement in system performance and robustness to system uncertainties, can be provided. The alleviation of the high-frequency switching or chattering due to traditional SMC [25] can be achieved. Comparing to the traditional PI-type control on piezoelectric positioning systems [1], [2], the faster response in trajectory tracking can be obtained simply. In addition, because the DSMC scheme is based on the equivalent dynamic model, the system stability and robustness also can be analyzed by mathematics. Using the proposed DSMC technique for trajectory tracking control of the piezoelectric positioning stage, the main advantages can be provided as follows. An asymptotical convergence of trajectory tracking can be obtained. The rate of convergence can be analyzed by mathematics and regulated by the dierent values of the control parameters. The chattering eect due to traditional slidingmode control can be alleviated. Robustness to an external disturbance-load can be obtained. To validate the proposed control design, implementation of a computer-based control and a one-dimensional piezoelectric positioning stage are carried out. Based on the implemented system, experimental results illustrate the feasibility of the trajectory tracking for scanning applications in microscopy.

Fig. 1. Equivalent model for dynamics of motion of a piezoelectric positioning stage.

hysteresis, Va indicates the applied voltage to piezoelectric positioning stage, m denotes the eective mass of the moving plate on the piezoelectric positioning stage, fL denotes the external load eect on the stage, k1 x + k2 x2 denotes the strain-dependent function, k1 , k2 , ke , kq , kq1 , kq2 , and kq3 denote the coecients related to the dynamics of motion of the piezoelectric actuator inside this stage. These expressions, shown in (1) and (2), can be viewed as the combination of a linear motion dynamics, nonlinear strain-dependent function, and the hysteretic relationship between a state variable q and excitation x [21], i.e., the use of (2) is equivalent to mainly generate the hysteresis eect in (1). Block diagram of the equivalent model consisting of (1) and (2) can be shown in Fig. 1. Moreover, dene the state variables in the following: x1 = x, x2 = x, x3 = Va , x4 = q, and dVa = u, dt

(3)

where x1 and x 1 denote the displacement and velocity of the moving plate, respectively; x3 denotes the applied voltage to the piezoelectric positioning stage; x4 can be viewed as the hysteresis eect; u is the input of the equivalent model. Substituting (3) into (1) and (2), the overall dynamics of the piezoelectric positioning stage can be expressed as follows: x 1 = x2 , k1 k2 ke kq fL x 2 = x1 x2 + x3 x4 , m m 1 m m m x 3 = u, x 4 = [kq2 |u| + kq3 u sgn (x4 )] x4 + kq| u.

II. System Dynamics Derivation To describe the dynamics of motion of the piezoelectric positioning stage implemented in this paper, the following dierential equations modied from the dynamics of motion of piezoelectric actuators [22] are given: d2 x + k1 x + k2 x2 + fL = ke Va kq q, dt2 dq dVa dVa dVa = kq1 kq2 q q kq3 q, dt dt dt dt m (1) (2)

(4)

where x denotes the output displacement of the piezoelectric positioning stage, q is dened as the state of the

From (4) it is observed that a fourth-order dierential equation for the dynamics of motion of the piezoelectric positioning stage is established. However, the state variable x4 , dened in (3), is unavailable for the piezoelectric positioning stage if the equivalent model (4) is used. This unavailability leads to a diculty in full-state feedback control design. Therefore, to overcome this problem, a reduced-order equivalent model based on (4) is further developed.

1874

ieee transactions on ultrasonics, ferroelectrics, and frequency control, vol. 53, no. 10, october 2006

Integrating the last equation of (4) and setting all the initial values to zero yields:
t

x4 (t) =
0

[kq2 |u( )| + kq3 u( ) sgn (x4 ( ))] x4 ( )d

t

+ kq1
0

u( )d

(5)

= kq1 x3 (t) (x4 (t), u(t)) , where (x4 (t), u(t))

t

kq2 |u( )| + kq3 u( ) sgn (x4 ( )) x4 ( )d.

0

Considering the state x4 (t) as a hysteretic state impacting on the rst two equations of (4), the third-order dynamic model with hysteresis eect can be expressed as follows: x 1 = x2 , x 2 = a1 x1 a2 x2 1 + a3 x3 fq (x3 , x4 , d) , x 3 = u, (6)

where a1 = k1 /m, a2 = k2 /m, and a3 = ke /m denote the system parameters related to the mechanical motion dynamics and fq (x3 , x4 , d) indicates a specic uncertain function that consists of the state x4 , external load force fL , and lumped parameter uncertainty d, as shown in the following: fq (x3 , x4 , d) = kq kq kq1 fL (x4 , u) x3 + + d, m m m

(7)

where d is due to the variations of the system parameters of a1 , a2 , a3 , and fL . To validate the developed model that has the frequency-dependent hysteresis behavior, the sinusoidal input voltage with frequencies of 0.5 and 1.0 Hz is applied to both the mathematic model (6) and the actual piezoelectric positioning stage. Figs. 2(a) and (b) illustrate the results of the simulation and experimental tests, respectively. The displacement output from the piezoelectric positioning stage in experiments is measured by a laser displacement interferometer and the parameters values used in simulations are given as follows: a1 = 10.5, a2 = 10, a3 = 1.014 106, kq1 = 2.961 106, kq2 = 2.292 107, kq3 = 2.865 109 , and m = 0.01 kg. All these values given above, except for the eective mass m, are experimentally chosen so that the hysteresis response of the proposed equivalent model agrees with that of the actual piezoelectric positioning stage. By comparing Figs. 2(a) and (b), we can conrm that the equivalent model (6) can be used to describe the dynamics of motion of the piezoelectric positioning stage with hysteresis behavior.
Fig. 2. Hysteresis responses from (a) the proposed mathematical model and (b) the actual piezoelectric positioning stage.

III. Dynamic Sliding-Mode Tracking Control In this section, the dynamic sliding-mode tracking control with backstepping is developed for the trajectory

shieh and huang: trajectory tracking using dynamic sliding-mode control

1875

tracking of the piezoelectric positioning stage. According to the developed dynamic model (6), the tracking errors between desired trajectories and real output displacements are dened as follows: e1 = x1 xm , e2 = x2 x2d , e3 = x3 x3d , (8) (9) (10)

where c1 is a suitable, positive-real constant. Then, the stabilizing control x3d associated with the rst sliding surface (17) can be designated as follows: 1 [x3d1 (t) + x3d2 (t)] , (19) a3 x3d1 (t) = d1 1 (t) c1 e2 (t) + a1 x1 (t) + a2 x2 2d (t), 1 (t) + x f x 3d2 (t) = x3d2 1 sgn (1 (t)) , (20) x3d (t) = where d1 is a suitable, positive-real constant, f is a small, time-constant of the rst-order, low-pass lter, 1 is chosen as the upper-bound of |fq | (i.e., 1 |fq |), and sgn() denotes the signum function dened as: +1 as z > 0 sgn(z ) = (21) 0 as z = 0. 1 as z < 0 Moreover, according to the sliding surface (18), the control input u for stabilizing the subsystem composed of (16) can be designated as follows: u = d2 2 1 2 sgn (2 ) , (22)

where xm denotes the desired trajectory; x2d and x3d are dened as the auxiliary destination states for x2 and x3 , respectively. According to the denitions of (8)(10) and the backstepping approach [26], the error dynamics for trajectory tracking can be derived step by step. For the rst error dynamics, the time-derivative of (8) can be derived as follows: e 1 = e2 + x2d x m. (11)

In (11), the auxiliary destination state x2d can be viewed as a stabilizing control [26] for the subsystem consisting of (11). Dening x2d = x m , the error dynamics (11) becomes: e 1 = e2 . Moreover, taking the time-derivative of (9) yields: 2 x 2d = a1 x1 a2 x2 2d . e 2 = x 1 + a3 x3 fq x (13) Substituting (10) into (13), the error dynamics for subsystem consisting of (13) can be derived as follows: e 2 = a3 e3 + a3 x3d a1 x1 a2 x2 2d . 1 fq x (14) Similar to x2d , the state x3d can be viewed as a stabilizing control for the subsystem consisting of (14). Furthermore, taking the time-derivative of (10) yields: e 3 = x 3 x 3d = u x 3d . (15) (12)

where d2 and 2 are suitable, positive-real constants and 2 = 21 |1 |. Block diagram of the overall piezoelectric positioning stage under the use of the proposed DSMC for trajectory tracking is shown in Fig. 3. In Fig. 3, the state-feedback block represents the rst equation of (20), and the stabilizing controller block represents (22). Consequently, the trajectory tracking of the piezoelectric positioning stage using the DSMC with the stable sliding surfaces is completed. IV. Analysis of System Stability and Performance To investigate the system stability of the proposed DSMC for trajectory tracking, the following steps for stability analysis are used. Step 1 The rst Lyapunov function is chosen as follows: V1 (1 ) = 1 2 . 2 1 (23)

Setting u = u +x 3d , where u is dened as an auxiliary control input for the proposed model, the error dynamics (15) becomes: e 3 = u . (16)

Observe that there is a mismatched uncertainty fq in the error dynamics consisting of (12), (14), and (16). This mismatched uncertainty often leads to the degraded control performance or instability for systems with the design of traditional PI or SMC controller. To improve the above drawbacks due to traditional controller design and to guarantee the robustness to the mismatched uncertainty, the DSMC with backstepping is developed in this study. Two asymptotically stable sliding surfaces of 1 (t) and 2 (t) rst are selected as follows: 1 (t) = c1 e1 (t) + e2 (t), 2 (t) = e3 (t), (17) (18)

Taking the time derivative of (23) yields: 1 = 1 V 1 = 1 (c1 e 1 + e 2) = 1 c1 e2 a1 x1 a2 x2 2d fq . 1 + a3 e3 + a3 x3d x (24) Substituting (19) into (24), the above equation can be rewritten as:
2 1 = d1 1 V + 1 (e3 + x3d2 fq ) .

(25)

1876

ieee transactions on ultrasonics, ferroelectrics, and frequency control, vol. 53, no. 10, october 2006

Fig. 3. Block diagram of the proposed DSMC scheme on the piezoelectric positioning stage.

Using the equation of x3d2 = f x 3d2 1 sgn (1 ) derived from the second equation of (20), the following result can be deduced: 1 d1 2 + 1 2 = (1 |1 | |1 | |fq |) 1 f x V 3d2 ,
1 2 d1 1 2 d1 1

From the Lyapunov stability theory of view, the boundednees of 1 (t) and 2 (t) is guaranteed if all the states and the control input of this system are bounded. Theorem 1 If the proposed DSMC is used in the trajectory tracking control of the piezoelectric positioning stage and if the values of the control parameters are chosen appropriately, the following results can be provided: The reaching condition [23] for sliding-mode control is satised; that is, 1 (t) = 2 (t) = 0 as t . Asymptotical stability for the tracking errors is obtained. The convergence rate of the tracking errors can be governed by the choice of the dierent values of the control parameters. Proof T Let S () = [1 2 ] . Selecting 1 ( S ) = (1/4) 2 2 2 2 1 2 + d 2 2 , 1 + 2 , 2 ( S ) = 1 + 2 , and W (s) = d 1 2 where d1 < d1 and d2 < d2 , the Lyapunov function chosen in (27) and its derivative satisfy the following properties: 1 ( S ) < V (S ) < 2 ( S ) , 2 2 V = d1 1 + d2 2 = W (S ) 0. (31) (32)

+ 1 2 |1 | (1 |fq |) 1 f x 3d2 , + 1 2 1 f x 3d2 . (26)

Step 2 The second Lyapunov function for the overall piezoelectric system is chosen as follows: 1 2 V (t) = V1 (t) + 2 (t). 2 Taking the time derivative of (27) yields:
2 =V 1 + 2 V 2 d1 1 + 1 2 1 f x 3d2 + 2 u . (28)

(27)

Substituting (22) into (28), the following result can be obtained: d1 2 d2 2 (2 |2 | |1 f x V 3d2 | |2 |) ,
1 2 d1 1

2 2 d2 2

|2 | (2 |1 f x 3d2 |) .

(29)

Moreover, from the second equation of (20), it can be found that the state x3d2 is ultimately convergent to +1 or 1 if sgn (1 ) is always positive or negative in a long period. This implies that the result of |f x 3d2 | 21 is held. Therefore, we choose 2 = 21 |1 | to cause the result of 2 = 21 |1 | |1 | |f x 3d2 | |1 f x 3d2 |. With the substitution of the preceding inequality, the equation shown in (29) becomes:
2 2 d1 1 V d2 2 .

In other words, the Lyapunov function with the property of (27) and (30) satises the LaSalle-Yoshizawa theorem [24] since 1 ( S ) and 2 ( S ) are two strictly increasing functions and W (S ) is a positive, denite function with respect to S . Consequently, the boundedness of 1 and 2 can be guaranteed. Moreover, due to the results of lim W (S ) = 0 and lim S ((t)) = 0, the trackingt t

(30)

error system consisting of (12), (14), and (16) is uniformly

shieh and huang: trajectory tracking using dynamic sliding-mode control

1877

(a)

(b)

(c)

(d) Fig. 4. (a) Machine structure and (b) working principle of the piezoelectric positioning stage, (c) implementation of the computer-based control, and (d) block diagram of the control system conguration.

Fig. 5. Block diagram of implementation of the traditional PI-type tracking control of the piezoelectric positioning stage.

1878

ieee transactions on ultrasonics, ferroelectrics, and frequency control, vol. 53, no. 10, october 2006

Fig. 7. Square trajectory response under the traditional PI control scheme: (a) tracking response, (b) applied voltage, and (c) tracking error.

guaranteed when the error states are trapped in the sliding surface of S ((t)) = 0. 2 From Theorem 1, the following characteristics can be concluded: the system stability is ensured, the robustness to mismatched uncertainty is provided, and the transient performance in trajectory tracking can be regulated by the choice of the control parameters values, d1 and d2 .

Fig. 6. Square trajectory response under the proposed DSMC scheme: (a) tracking response, (b) applied voltage, (c) tracking error, and (d) output of the sliding function.

asymptotically stable. Furthermore, solving the dierential equation of (30) with assumption of the initial (t0 ), the following equation can be obtained: S ((t)) exp [D (t t0 )] S ( (t0 )) , where D = (33)

V. Experimental Results To validate the proposed control approach, implementation of a computer-based control system with the piezoelectric positioning stage, Model PX300 (manufactured by Piezosystem Jena GmbH, Jena, Germany) is carried out. The machine structure of the implemented piezoelectric positioning stage is shown in Fig. 4(a). In Fig. 4(a), a scanned object is mounted on the top plate by two threaded diagonal holes to emulate the situation that a scanned material is located on the piezoelectric positioning stage in practice, and the stage is xed to a heavy base by using two diagonal holes. As shown in Fig. 4(b), the movement of the top plate on this stage is due to a levered transmission of extension of the stacked piezoactuators, which are based on the lead zirconate titanate (PZT) crystals,

2d1 0 . Because the matrix D has 0 2d two negative eigenvalues, the exponential convergence of S ((t)) to zero is guaranteed. Therefore, from (33) it can be found that the rate of convergence can be determined by the choice of the control parameters values, d1 and d2 ; that is, the transient performance of the trajectory tracking can be regulated. In addition, due to the use of the asymptotically stable sliding surfaces, shown in (17) and (18), the asymptotical stability for tracking errors can be

shieh and huang: trajectory tracking using dynamic sliding-mode control

1879

Fig. 9. Sinusoidal trajectory response under the traditional PI control scheme: (a) tracking response, (b) applied voltage, and (c) tracking error.

Fig. 8. Sinusoidal trajectory response under the proposed DSMC scheme: (a) tracking response, (b) applied voltage, (c) tracking error, and (d) output of the sliding function.

with solid state exure hinges. When an input voltage is applied to the stage, the extension of the stacked piezoactuators occurs, then this extension causes the bending of the exure mounts. Consequently, the movement of the stage occurs due to the deviation of the top plate. Moreover, the block diagram of the piezoelectric system conguration for experimentation is shown in Figs. 4(c) and (d). In Fig. 4(c), a data-acquisition interface card integrated into a PC is to mainly operate two functions: analog-todigital (A-to-D) conversion with input range 5 V and 12-bit resolution, and digital-to-analog (D-to-A) conversion with output range 10 V and 16-bit resolution. The main purpose of the amplier circuit realized by operational ampliers is to obtain a higher resolution in output displacement. Furthermore, in Fig. 4(d) the transfer function (380.25/s2 + 39s + 380.25) with rising time = 0.2 s is

chosen as the prelter to avoid exciting the unmodeled uncertainties in the piezoelectric stage when the step trajectory commands are given. In addition, because of a builtin low-pass lter with maximum bandwidth 10 Hz, the bandwidth of the closed-loop positioning system in this experimentation is restricted to several hertz or less. This restriction leads to problems of that the maximum rate of change of the desired trajectories must be limited to several hertz below. The rated values of the primary parameters about the piezoelectric positioning stage investigated in this study are listed as follows: motion distance = 0 240 m, input voltage = 0120 V, minimum displacement resolution = 1 nm, resonant frequency = 150 Hz, capacitance = 2300 nF, repeatability = 0.03%, and displacement sensor resolution = 41.67 mV/m. In this positioning system, the applied voltage is outputted by the D-to-A converter, then amplied through a linear amplier with 10-multiple gain around. As a result, due to strain of the stacked piezoactuators inside this stage, the moving plate on the stage is driven to move to a xed point. Simultaneously, the displacement signal from the signal conditioning module, which resolves the dierential capacitance from the capacitive displacement sensor, is outputted to the amplier circuit with the function of transforming the displacement resolution into

1880

ieee transactions on ultrasonics, ferroelectrics, and frequency control, vol. 53, no. 10, october 2006

Fig. 11. Triangular trajectory response under the traditional PI control scheme: (a) tracking response, (b) applied voltage, and (c) tracking error.

Fig. 10. Triangular trajectory response under the proposed DSMC scheme: (a) tracking response, (b) applied voltage, (c) tracking error, and (d) output of the sliding function.

83.34 mV/m. Then, the output signal from the amplier circuit is sampled by the A-to-D converter interface integrated in the computer. The values of the control parameters used in experimentation are given as follows: c1 = 7.8, d1 = 9.0, d2 = 11.0, f = 1.5, 1 = 1.0, and 2 = 2.0. All these values are purposely chosen to meet the requirement of both the high-performance tracking response and system stability. Moreover, for the convenience of the comparison, the PI controller-based piezoelectric positioning stage is implemented and the corresponding block diagram is shown in Fig. 5. First, the square trajectory command with magnitude of 050 m and frequency of 1 Hz is given and the corresponding responses under the cases of using the proposed DSMC and PI controller (kP = 2 and kI = 12) are shown in Figs. 6 and 7, respectively. Then, the sinu-

soidal and triangular trajectory commands with magnitude of 050 m and frequency of 1 Hz are given, respectively, and the corresponding responses under the cases of using the proposed DSMC and PI controller (kP = 2 and kI = 12) are shown in Figs. 811. From Figs. 6 11 we can nd that the piezoelectric positioning stage using the proposed DSMC has an improvement in transient response and tracking error for square, sinusoidal, and triangular trajectory-tracking response. This improvement implies that the piezoelectric positioning stage with the DSMC approach can be used to the trajectory tracking with high-performance requirement. Moreover, to show an excellent control performance of the proposed DSMC, the tracking response to the irregular trajectory command with the form of 12.5 [sin(4 t) + cos(2 t)] is illustrated in Fig. 12. From Fig. 12 it can be found that the system performance in irregular trajectory tracking also is good enough. This result shows that the DSMC-based piezoelectric positioning stage can be applied to the precise tracking system with the need of various trajectories scanning. Moreover, by choosing two dierent values of the control parameters, (d1 , d2 ) = (6, 7) and (d1 , d2 ) = (9, 10), the transient tracking response to the step trajectory is illustrated in Fig. 13. From Fig. 13 it can be conrmed that the convergence rate of the tracking error for step trajec-

shieh and huang: trajectory tracking using dynamic sliding-mode control

1881

Fig. 12. Irregular trajectory response under the proposed DSMC scheme: (a) tracking response, (b) applied voltage, and (c) tracking error. Fig. 14. Square trajectory-tracking response and its tracking error for the piezo-stage with an external disturbance load of 10 N by using (a) the DSMC and (b) the PI control.

tory response can be regulated by the choice of the control parameter values. In addition, to investigate the robustness of the proposed DSMC scheme, experimental results of the square trajectory-tracking responses with the condition of an external, disturbance load of 10 N on the stage are provided. Figs. 14(a) and (b) illustrate the results of both using the DSMC and using the traditional PI control scheme in square trajectory tracking of the piezoelectric stage. From comparison of Figs. 14(a) and (b), it can be observed that robustness of the piezoelectric stage to an external disturbance load can be substantially provided by using the proposed DSMC scheme.

VI. Conclusions The trajectory tracking of a piezoelectric positioning stage using a DSMC scheme is proposed in this paper. To completely describe the dynamics of motion of the piezoelectric positioning stage, an equivalent mathematical model constructed by adding the hysteresis and straindependent functions to a second-order, linear model of

Fig. 13. Transient in step trajectory tracking using dierent values of control parameters: (a) (d1 , d2 ) = (6, 7) and (b) (d1 , d2 ) = (9, 10).

1882

ieee transactions on ultrasonics, ferroelectrics, and frequency control, vol. 53, no. 10, october 2006 [15] B. Cannas, S. Cincotti, and I. Daneri, A generalization of a piece-wise linear circuit model of hysteresis, IEEE Trans. Magn., vol. 38, no. 2, pp. 901904, 2002. [16] S. Cincotti and I. Daneri, A PWL circuit approach to the denition of a -approximation model of scalar static hysteresis, IEEE Trans. Circuits Syst. I, vol. 49, no. 9, pp. 12901308, 2002. [17] S. Cincotti, Dynamic properties of a piece-wise linear circuit model of hysteresis, IEEE Trans. Magn., vol. 37, no. 5, pp. 33203323, 2001. [18] M. Goldfarb and N. Celanovic, Modeling piezoelectric stack actuators for control of micromanipulation, IEEE Contr. Syst. Mag., vol. 17, no. 3, pp. 6979, June 1997. [19] J. J. Tzen, S. L. Jeng, and W. H. Chieng, Modeling of piezoelectric actuator for compensation and controller design, Precision Eng., vol. 27, pp. 7086, Jan. 2003. [20] H. J. M. Adriaens, W. L. de Koning, and R. Banning, Modeling piezoelectric actuators, IEEE/ASME Trans. Mechatron., vol. 5, pp. 331341, 2000. [21] T. S. Low and W. Guo, Modeling of a three-layer piezoelectric bimorph beam with hysteresis, IEEE J. Microelectromech. Syst., vol. 4, pp. 230237, Dec. 1995. [22] S. E. Lyshevski, MEMS and NEMS: Systems, Devices, and Structures. New York: CRC Press, 2002. [23] V. I. Utkin, Sliding Modes in Control and Optimization. New York: Springer-Verlag, 1992. [24] H. K. Khalil, Nonlinear Systems. Upper Saddle River, NJ: Prentice-Hall, 2002. [25] S. H. Zak, Systems and Control. New York: Oxford Univ. Press, 2003. [26] M. Krstic, I. Kanellakopoulos, and P. Kokotovic, Nonlinear and Adaptive Control Design. New York: Wiley, 1995.

motion of a mechanical system is proposed. To validate the mathematical model, the hysteresis response from the actual piezoelectric positioning stage and the mathematical model, respectively, under sinusoidal voltage input with frequency of 0.5 and 1.0 Hz are illustrated. Moreover, according to the mathematical model, the DSMC with asymptotically stable sliding surfaces is proposed for trajectory tracking of the piezoelectric positioning stage. In the DSMC approach, an augmented, rst-order lter is used to avoid the direct dierentiation in discontinuity and to reduce the chattering phenomenon introduced by the traditional sliding-mode control. Using this proposed control approach, the following advantages can be provided for the piezoelectric positioning stage: an asymptotical convergence in trajectory tracking can be obtained, the rate of convergence can be analyzed by mathematics and regulated by the choice of values of the control parameters, the chattering eect due to the traditional sliding-mode control can be alleviated, and robustness to an external disturbance load can be obtained. And, experimental results are illustrated to show the validity of the proposed control approach for practical trajectory-tracking applications. References
[1] J. Stroscio and K. Kaiser, Scanning Tunneling Microscopy. New York: Academic, 1993. [2] M. A. Paesler and P. J. Moyer, Near-Field Optics: Theory, Instrumentation, and Applications. New York: Wiley, 1996. [3] J. P. Fillard, Near Field Optics and Nanoscopy. Hackensack, NJ: World Scientic Publ., 1996. [4] A. J. Fleming and S. O. Reza Moheimani, Control orientated synthesis of high-performance piezoelectric shunt impedances for structural vibration control, IEEE Trans. Contr. Syst. Technol., vol. 13, pp. 98112, Jan. 2005. [5] J. W. Gardner, V. K. Varadan, and O. O. Awadelkarim, Microsensors, MEMS, and Smart Devices. New York: Wiley, 2001. [6] R. L. Clark, W. R. Saunders, and G. P. Gibbs, Adaptive Structures: Dynamics and Control. New York: Wiley, 1998. [7] S. B. Jung and S. W. Kim, Improvement of scanning accuracy of PZT piezoelectric actuator by feed-forward model-reference control, Precision Eng., vol. 16, no. 1, pp. 4955, 1994. [8] P. Ge and M. Jouaneh, Tracking control of a piezoceramic actuator, IEEE Trans. Contr. Syst. Technol., vol. 4, pp. 209216, 1996. [9] G. S. Choi, H. S. Kim, and G. H. Choi, A study on position control of piezoelectric actuators, in Proc. IEEE Int. Symp. Ind. Electron., 1997, pp. 851855. [10] C. L. Hwang, C. Jan, and Y. H. Chen, Piezomechanics using intelligent variable-structure control, IEEE Trans. Ind. Electron., vol. 48, pp. 4759, Feb. 2001. [11] Y. Bernard, E. Mendes, and F. Bouillault, Dynamic hysteresis modeling based on Preisach model, IEEE Trans. Magn., vol. 38, pp. 885888, 2002. [12] G. Bertotti, Dynamic generalization of the scalar Preisach model of hysteresis, IEEE Trans. Magn., vol. 28, no. 5, pp. 25992601, 1992. [13] Y. Bernard, E. Mendes, and F. Bouillault, Dynamic hysteresis modeling based on Preisach model, IEEE Trans. Magn., vol. 38, no. 2, pp. 885888, 2002. [14] D. Song and C. J. Li, Modeling of piezo actuators nonlinear and frequency dependent dynamics, Mechatronics, vol. 9, no. 4, pp. 391410, 1999.

Hsin-Jang Shieh (M02) received the B.S. and Ph.D. degrees in electrical engineering from the National Central University, ChungLi, Taiwan, R.O.C., in 1992 and 1997, respectively. From 1997 to 2002, he was with the Mechanical Industry Research Laboratories, Industrial Technology Research Institute (ITRI), Hsinchu, Taiwan, as a researcher of the Micro-Electro-Mechanical Systems (MEMS) Division. Since August 2002, he has been with the National Dong Hwa University, Hualien, Taiwan, where he is currently an assistant professor in the Department of Electrical Engineering. His research interests include design and control of piezoelectric mechanisms, power converters, photovoltaic systems, and control theory applications.

Po-Kai Huang was born in Taipei, Taiwan, R.O.C., in 1979. He received the B.S. and M.S. degrees in electrical engineering from the Chung Yuan Christian University, Chung-Li, Taiwan, R.O.C., in 2001 and 2003, respectively. He is currently working toward the Ph.D. degree in the Department of Electrical Engineering, National Dong Hwa University, Hualien, Taiwan, R.O.C. His research interests include AC motor drives, articial intelligence control, and piezoelectric mechanisms control.