This action might not be possible to undo. Are you sure you want to continue?

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 50, NO. 11, NOVEMBER 2005

**Quasi-Continuous High-Order Sliding-Mode Controllers
**

Arie Levant

Abstract—A universal ﬁnite-time-convergent controller is developed capable to control the output of any uncertain single-input-single-output system with a known permanent relative degree . The tracking error is _ ... steered to zero by means of a control dependent only on and continuous everywhere except the set = _ = = = 0. A robust output-feedback controller version provides for the tracking accuracy proportional to the sampling noise magnitude. Index Terms—Finite-time stability, high-order sliding mode, output feedback control, robustness.

on the homogeneity reasoning [18], [19], and can be considered as a demonstration of the principles [18]. Combining with the recently proposed robust exact ﬁnite-time-convergent differentiator [16] an output-feedback controller is obtained providing for exact tracking 0 if the measurements of the tracking error are exact, and for proportional to the maximal measurement error otherwise. Its transient features are much better than those of the known r -sliding controllers [15], [16] (Section VI). Simulation demonstrates the practical applicability of the new controller. II. PRELIMINARIES AND THE PROBLEM STATEMENT Consider a smooth dynamic system with a smooth output function , and let the system be closed by some possibly-dynamical discontinuous feedback and be understood in the Filippov sense [5]. Then, provided (r 01) that successive total time derivatives ; ; _ ...; are continuous functions of the closed-system state-space variables, and the set (1) is a nonempty integral set, the motion (1) is called r -sliding (r th-order sliding) mode [13], [3], [16]. The standard sliding mode used in the most variable structure systems is of the ﬁrst-order ( is continuous, _ is discontinuous). and Consider a dynamic system of the form

I. INTRODUCTION Sliding mode control remains one of the most robust and effective tools to cope with heavy uncertainty conditions [22]. The main drawback of the standard sliding modes is mostly related to the so-called chattering effect caused by the high-frequency control switching [7], [8]. Let s be the output variable of an uncertain single-input–singleoutput (SISO) dynamic system and w(t) be an unknown-in-advance smooth input, both available in real time. The task is to establish and keep = s 0 w(t) = 0. The standard sliding-mode control u = 0k sign is applicable if the relative degree is 1, i.e., if _ 0 explicitly depends on the control u, and _ u > 0. Higher-order sliding mode [13], [16] is applicable for controlling SISO uncertain systems of arbitrary relative degrees [3], [6], [11], [14]–[16], [20]. The corresponding ﬁnite-time-convergent controllers (r -sliding controllers) [13], [16] require actually only the knowledge of the system relative degree r . The produced control is a discontinuous function of the tracking error and of its real-time-calculated successive derivatives (r01) . The accuracy is improved in the presence of _ ; . . . ; ; ; switching delays, and the chattering effect is successfully treated, provided the control derivative is used as a new control input [3], [13]. The discontinuity set of controllers [15], [16] is a stratiﬁed union of manifolds with codimension varying in the range from 1 to r , which causes certain transient chattering. To avoid the chattering one needs to increase artiﬁcially the relative degree r , inevitably complicating the controller implementation [15], [16]. The ﬁnite-time-stable exact tracking is lost with alternative controllers developed in [2] and [21] for r = 3 and r = 2, respectively. A sliding-mode controller of a new type is proposed in this note, (r 01) being a feedback function of ; ; _ ; . . . ; , continuous everywhere except the manifold deﬁned by the equations

(r 01) = _ = = 111 = =0

x _ = a(t; x) + b(t; x)u = (t; x):

(2)

Here, x 2 Rn ; a; b and : R n+1 ! R are unknown smooth functions, u 2 R; n is also uncertain. The task is to provide in ﬁnite time for exact keeping of 0. The relative degree r of the system is assumed to be constant and known. In other words [10], the control explicitly appears ﬁrst time in the r th total time derivative of and

(r) = h(t; x) + g(t; x)u

(3)

**where h(t; x) = (r) ju=0 ; g (t; x) = (@=@u) (r) = 6 0. It is supposed that for some Km ; KM ; C > 0
**

0

< Km

@ (r) KM @u

j r j u j C

( ) =0

(4)

(1)

of the r -sliding mode. The mode 0 is established after a ﬁnite-time transient. In the presence of errors in evaluation of the output and its derivatives, a motion in some vicinity of (1) takes place. Therefore, control is practically a continuous function of time, for the trajectory never hits the manifold (1) with r > 1. The controller design is based

which is always true at least locally. Trajectories of (2) are assumed inﬁnitely extendible in time for any Lebesgue-measurable bounded control u(t; x). Finite time stabilization of smooth systems at an equilibrium point by means of continuous control is considered in [1]–[9]. In our case, (r 01) ) providing for 0, _ ...; any continuous control u = U (; ; would satisfy the equality U (0; 0; . . . ; 0) = 0h(t; x)=g (t; x), whenever (1) holds. Since the problem uncertainty prevents it [18], the control has to be discontinuous at least on the set (1). Hence, the r -sliding mode = 0 is to be established. The controller which is designed in this note, is r -sliding homogeneous [18], which means that the identity r r01 ; (r 01) (r 01) _ . . . ; ) is kept for any U (; ; _ ...; ) U ( ; > 0. III. FEEDBACK DESIGN As follows from (3) and (4)

Manuscript received April 9, 2004; revised November 19, 2004 and May 9, 2005. Recommended by Associate Editor J. Berg. The main results of this note were presented at the Conference on Decision and Control 2003. The author is with the School of Mathematical Sciences, Tel-Aviv University, 69978 Tel-Aviv, Israel (e-mail: levant@post.tau.ac.il). Digital Object Identiﬁer 10.1109/TAC.2005.858646

(r) 2 [0C; C ] + [Km ; KM ]u:

(5)

The closed differential inclusion is understood here in the Filippov sense [5], which means that the right-hand vector set is enlarged in

0018-9286/$20.00 © 2005 IEEE

UNIVERSAL OUTPUT-FEEDBACK SISO CONTROLLER Controller (5) requires the real-time exact calculation or direct measurement of . r 0 1. . IV.r = N = jj 9 = ' =N = sign = + . in order to satisfy certain convexity and semicontinuity conditions. KM . . This inclusion does not “remember” anything on system (2) except the constants r. . . . . Denote constants i the inequalities j (i) j are established in ﬁnite time for any " "( 0 ) . . which allows the real-time robust (r 0 1)th-order differentiation of [16]. . r 0 1. 11. Equality (3) implies j(r) j C +KM . . NOVEMBER 2005 1813 a special way [18]. Let i = 0. (r01) .IEEE TRANSACTIONS ON AUTOMATIC CONTROL. the ﬁnite-time stabilization of (5) at the origin solves the stated problem simultaneously for all systems (2) satisfying (4). Km . i r i =r = 0. Consider uncertain system (2). . Thus.r i. NO. [18] achieve '0 ' N where . VOL. _ . C.r i. Other differentiators [8] can be also used. i > 0. . Combining controller (6) and the homogeneous differentiator [16]. . 50. (4).

N 00 0 9 0 = j j + .

. N 00 0 9 = ' =N 0. .r j _ .r 0.r = 0 i = i. . Ni. r 0 1. The inequality j(9 Ni.r i+1) i.r is positive deﬁnite. . it can be redeﬁned by continuity) except the point = _ = 1 1 1 = (i01) = 0. i01) ) is 1 holds whenever Ni. Theorem 1: Provided . The following proposition is easily proved by induction.r (i) i (r i i)=(r 1.r 0. . .r i. Proposition 1: Let i = 0. The function 9i.r values of the control on any set of the zero Lebesgue measure do not inﬂuence the solutions.r (i) i (r i i)=(r 1.. .r i.r > 0.r i+1) i 1.r (.e. . _ = 1 1 1 = (i) = 0.. i.e.r 0. continuous everywhere (i.

. . .1 . . .

> 0 are chosen sufﬁciently large in the list order.r01 . the controller .

1 . . . . . .

obviously 'i.r = (i) + 0 1=(r 0i+1) 'i01.r01 are positive numbers. Recall that according to the Filippov deﬁnition .r .

.i Ni01. z1 . . Any time the ﬁnite-time stability is mentioned in this note it means that the maximal possible transient time is a locally bounded function of initial conditions [18]. The proof is given in Section V. _ . . .r r r =r r =r k k k r k = r k k k k r k = r k k k r r r u = 0 9 01 r .r . Each choice of parameters . . (6). z 01 ) (7) 1 ( 01) sign(z0 0 ) + z1 (8) z _0 = v0 . . . v = 0 0 L 1 ( 0 ) jz 0 v 01j( 0 01) ( 0 )sign(z 0 v 01) + z +1 k = 1. .r u = 09 01 (z0 . The ﬁnite-time stable r -sliding mode 0 is established in the system (2). . r 0 2 (9) z _ 01 = 01 Lsign(z 01 0 v 02 ) (10) r . . ( 01) r (6) is r -sliding homogeneous and provides for the ﬁnite-time stability of (5). v0 = 0 L jz0 0 j z _ = v . It follows from Proposition 1 that control (6) is continuous everywhere except the r-sliding mode = _ = 1 1 1 = (r01) = 0. . (6). .

. .1 . . . .

Km . KM . most conveniently by computer simulation. Km . The parameter is chosen speciﬁcally for any ﬁxed C. Following are controllers with r 4 and simulation-tested . is to be negative with (@=@u)(r) < 0. Obviously.r01 determines a controller family applicable to all systems (2) of the relative degree r . KM . avoiding redundantly large estimations of C.

PROOF OF THEOREM 1 The proof is based on a few Lemmas. which corresponds to the r -sliding homogeneity [15]. . . . Assign the weights (homogeneity degrees) r 0 i to (i) . . i = 0. are established in ﬁnite time for some positive i . 6 = 8 can be taken. Lemma 2: Let 1 i r 0 2. Theorem 5: Let > 0 be the constant sampling interval and the noises be absent. L is to satisfy L The sequence i is chosen in advance [16]. The computer-tested values 1 = 1:1. (i) ) of the Each homogeneous locally-bounded function ! (. only one parameter is really needed to be tuned.4 3=4 =4 2=3 01=2 0:5jj sign)][j j + (j _ j + 0:5jj ) ] . . . are established in ﬁnite time for some positive constants i . The lacking values need to be tuned in the unlikely case of r > 6. i = 0. . these values are sufﬁcient for up to the ﬁfth-order differentiation and r 6. . . r 0 1 and the weight 1 (minus system homogeneity degree [1]) to t. _ . Ni. Lemma 1: The weight of Ni. Then with properly chosen parameters of the controller (7)–(10) the inequalities j (i) j i "(r0i)=r . Due to the recursive form of the differentiator. Hence. which implies that !=Ni. . 3 = 2. i = 0. 5 = 5.i : 1) u = 0sign . both L and are found by computer simulation. . Usually. r 0 1. everywhere. i = 0. = + 3[ + (j _j 3 + 0:5jj3=4)01=3(_ + 4) '3. Only the main proof points are listed below. therefore.r for some c > 0. r 0 1. 2 = 1:5. . . With exact measurements (" = 0) the r -sliding mode 0 is established in the closed system. KM are known. .r equals r 0 i. V. . where the parameters of the differentiator (8)–(10) are chosen acC + KM . 2=3 2) u = 0( 3 01=2 + 2(j _ j + jj ) (_ + jj sign)]=[j j + 3) u = 0[ 2(j _ j + jj2=3)1=2 ].r is a positive–deﬁnite locally bounded function (Proposition 1).r is bounded on a unit sphere and. 4 = 3. then for any positive . . . cording to the condition j (r) j L. Then inequalities j (i) j i r0i . . in the case when C and Km . jj ". r 0 1. weight r 0 i satisﬁes the inequality j! j cNi. Indeed. _ + jj1=2 sign2 )= =(j _ j + jj1=2). globally attracting trajectories in ﬁnite time. . Theorem 4: Let be measured with a Lebesgue-measurable noise . .

i . i . i+1 with sufﬁciently large .

i+1 > 0 the inequality (r 0i01)=(r 0i) (r 0i01)=(r 0i) j(i+1) + .

r 9i.r j i+1 Ni.r provides for the ﬁnite-time establishment and keeping of the inequality r 0i)=(r 0i+1) r 0i)=(r 0i+1) j(i) + .i+1 Ni.

< i =(3. g for some ﬁxed > 0. .i Ni(0 9i01.r j i Ni(0 .r 1.r j Proof: Consider the point set ( ) = f(. (i) )jj9i. . . 1. .r _ .

r j implies j(i) j 2. The inequality r 0i)=(r 0i+1) j9i.i ). < 1=3.

i Ni(0 and. therefore.r () 1 (). 1. where 1 () is deﬁned by the inequality ) ( 0 +1) ( ) + .

.r i i . It may have inﬁnite derivatives when certain surfaces are crossed. . ti+1 ) (the discrete sampling case). . 3=4 2=3 1=2 The control is a continuous function of time everywhere except the r -sliding set (1). r01 . 3[j j + (j _ j + 0:5jj ) ] . . . Theorem 2: Let the control value be updated at the moments ti . . . . N (00 9 01 1 i i r i i = r . t 2 [ti . The following result shows robustness of controller (6) with respect to measurement errors. Then. Then with some positive 3. . i = 1. . . r 0 1. j _ j < 1 r01 . . with ti+1 0 ti = = const > 0. j(r01) j < r01 with some positive constants 0 . controller (6) provides in ﬁnite time for keeping the inequalities jj < 0 r .r N3 4 = j j + u = 0'3 4 =N3 4 . 1 . . . . Theorem 3: Let (i) be measured with accuracy i "(r0i)=r for some ﬁxed i > 0. That is the best possible accuracy attainable with discontinuous (r) [13]. All further theorems are standard consequences [18] of the r -sliding homogeneity of controller (6) and Theorem 1.

N 00 i (r i i)=(r 1.r 0i+1) : .

. . where 0 . (i01) ) = w 0 r 0 r +1 (0r 0i+1) (i01) . in its turn. is deﬁned above. . Rehomogeneous functions of . . Any function deﬁned on the homogeneous sphere = 1 is uniquely extended to the function 8 of the weight w > 0 deﬁned in the _ . . + are _ . . ).. NOVEMBER 2005 That is equivalent. where p = p=r + achieve some continuous on the sphere functions 10 and 1+ . . Thus. . respectively. . _ p=(r01) . VOL. . . so that ( ) 2 = f(. whole space . (i01) by the formula 8(. smooth everywhere except 0. 50. NO. . . Functions 10 and 1+ can be approximated on the sphere by some smooth functions 20 and 2+ from beneath and from above. functions 20 and 2+ are extended by homogeneity to the continuous homogeneous functions 80 and 8+ of . _ . 11. where the function ( . .. . . . . . _ . _ . to 0 (i) + . (i01) of the weight r 0 i. . + ((i01))p=(r0i+1) . . (i01) of the weight r 0 i. stricting 0 and + to the homogeneous sphere of the radius = 1. (i01) )j80 (i) 8+ g. . . _ .1814 IEEE TRANSACTIONS ON AUTOMATIC CONTROL. Prove now that 2 is invariant and attracts the trajectories with large . p = 2r!.

Kinematic car model. achieve differentiating that with sufﬁciently large Fig. The inequality j9i. Suppose that at the initial moment + > 0 and. 1.r _ + is homogeneous of the weight r 0i01 and.r . The “upper” boundary of 2 is given by the equation + = (i) 0 8+ = 0.r j is assured outside of 2 . j8 for some . 9i. . Taking into account that 8 (r 0i01)=(r 0i) _ + j Ni. and j+ j 1 Ni. therefore.i+1 . .r according to Lemma 1. 1 > 0.

+1 i _+ _ + (0.

+1 + +1 )N ( 0 01) ( 0 ) 0 8 ( 0 01) ( 0 ) (0.

+1 + +1 + )N 01 + ( 0 01) ( 0 ) : (0.

+ vanishes in ﬁnite time with . +1 + +1 + ) 1 i i i i i i r i.r = r r i i = r i Hence.r i = r i i r i.

2 is invariant. the trajectory inevitably enters the region 2 in ﬁnite time. therefore. Obviously. 9i. Similarly. Thus.r 0 . the trajectory enters 2 if the initial value of + is negative and. Choosing 80 and 8+ sufﬁciently close to 0 and + on the homogeneous sphere and .i+1 large enough.

i+1 large enough. achieve from Lemma 1 that 2 1 ( i =(3.

Since N0.r = j j. _ + .r = . Lemma 2 is replaced by the next simple lemma with i = 0. '0.i )) and the statement of Lemma 2.

1 jj(r01)=r signj 1 jj(r01g=r Lemma 3: The inequality j provides with 0 1 < .

1 for the establishment in ﬁnite time and keeping of the identity 0. The proof of the theorem is now ﬁnished by the similar to Lemma 2 proof that for any > 0 with sufﬁciently large the inequality =2 1=2 j(r01) + .

3-sliding tracking errors. in order to provide some time for the differentiator convergence. The car trajectory. SIMULATION EXAMPLE: CAR CONTROL Consider a simple kinematic model of car control Fig. but is much worse with sampling noises). where g (x) and y are assumed to be available in real time. In the _j absence of noises the tracking accuracies j j 5:4 1 1007 . but its transient features some chattering (Fig. 2). respectively. j j 0:042 were obtained. 3-sliding car control [15]. Let v = const = 10 m/s. in order to provide for better performance in the presence of measurement errors (L = 25 is also sufﬁcient.r is established in ﬁnite time and kept afterwards. l = 5 m. VI. The relative degree of the system is 3 and the 3-sliding controller [15]. 3(a)–(d). [14]. L = 400. steering angle and its derivative u are shown in Fig. Deﬁne = y 0 g (x). which mainly corresponds to the asymptotics stated in Theorem 5. [16] solves the problem. g (x) = 10 sin(0:05x) + 5. It was taken = 1.r01 Nr10 2. 1). The task is to steer the car from a given initial position to the trajectory y = g (x). x = y = ' = = 0 at t = 0. With = 1005 the accuracies jj 5:6 1 10010 . j 2:4 1 1004 . The integration was carried out according to the Euler method (the only reliable integration method with discontinuous dynamics). The control was applied only from t = 1. 2. ' is the orientation angle. The differentiator parameter L = 400 is taken deliberately large. 3(c) that the control u remains continuous until the entrance into .r 9r 02. It is seen from Fig.r j Nr 02. Also The controller parameter is convenient to ﬁnd by simulation [3]. 3-sliding controller N 3 can be applied here. l is the length between the two axles and is the steering angle (Fig. the sampling step being equal to the integration step = 1004 . j _ j 1:4 1 1005 . The resulting output-feedback controller (8)–(10) is u = 0[z2 + 2(jz1 j + jz0 j2 3 )01 2 (z1 + jz0 j2 3 signz0 )]=[jz2 j + 2(jz1 j + jz0 j2 3 )1 2 ] z _0 = v0 v0 = 014:7361jz0 0 j2 3 sign(z0 0 ) + z1 z _1 = v1 v1 = 030jz1 0 v0 j1 2 sign(z1 0 v0 ) + z2 z _2 = 0440 sign(z2 0 v1 ): = = = = = = = x _ = v cos ' y _ = v sin ' ' _ = v=l tan _ = u where x and y are Cartesian coordinates of the rear-axle middle point. j j 0:0042 were attained. v is the longitudinal velocity.

11. Fig. tracking accuracies j j 0:02. New 3-sliding car control. 50. which is also quite feasible. 4(c) that the control u is a continuous function of t. The steering angle remains rather smooth and is quite feasible. 2 and 3). NOVEMBER 2005 1815 Fig.1 the accuracies _ j 0:62. Simulation shows that the previous controller [15]. VOL. The steering angle vibrations have the magnitude of about 7 and the frequency 1. The performance does not change. It is seen from Fig. NO. It is also a sliding-mode SISO controller of a new type. It is actually only the second known family of such controllers.1 m is shown in Fig. j j 1:3 were obtained. the 3-sliding mode.01 m the _ j 0:14. [16] is also much more sensitive to the parameter choice and noises. 4. when the frequency of the noise varies in the range 100–100 000. 4. j the asymptotics stated by Theorem 4. The performance of the controller with the measurement error magnitude 0.IEEE TRANSACTIONS ON AUTOMATIC CONTROL. CONCLUSION A new arbitrary-order sliding mode controller is proposed. VII. Performance with the input noise magnitude 0. j j 2:8 which corresponds to changed to j j 0:20. In the presence of output noise with the magnitude 0. for it provides for the ﬁnite-time stable sliding motion on the zero-dynamics manifold of high .1 m. j With the measurement noise of the magnitude 0. The advantages of the new controller are obvious (compare Figs. 3.

vol. vol. the indirect one allows for larger initial conditions in the frequency estimate. [9] Y. K. 875–892. J. 2002. 2003. output tracking with multiple sinusoidal disturbance rejection is achieved for linear minimum phase systems and in [19] an adaptive extension is presented when the system parameters are unknown. Only boundedness of the noise is needed. vol. London. the problem of adaptive compensation of sinusoidal disturbances has attracted a considerable attention. Nonlinear Control Systems. 2005. “Higher-order sliding modes. pp. no. 751–766. The authors are with the Dipartimento di Ingegneria Elettronica. 6. 614–619.. [20] H. Two schemes (a direct one and an indirect one) are presented and analyzed in [1]: While the direct scheme is local. NV. Maui. “Homogeneity approach to high-order sliding mode design. 860–865. 9/10. 76. J. 41.: Springer-Verlag. Control. “Finite-time stabilization and stabilizability of a class of controllable systems. Orlov.” Syst. 5. 1989.” IEEE Trans. A. New York: SpringerVerlag.2005. [4] S. “Finite time stability and robust control synthesis of uncertain switched systems. Dec.. Both the proposed controller and its output-feedback version are very robust with respect to measurement noises. Dyna. S.” in Proc. while the output regulation problem is addressed in [27] for a class of large-scale nonlinear interconnected systems perturbed by a neutrally stable exosystem via a decentralized error feedback controller. no. Università di Roma “Tor Vergata. vol. Isidori..uniroma2. Utkin. Goh. Barbot and M. Control Syst. pp. J. Recently. [11] M. 823–830. Differential Equations with Discontinuous Right-Hand Side.” SIAM J. vol. 6. 2001. 5. pp. P. I. E. global output feedback regulators for the same class of nonlinear systems considered in this note under the MP assumption have been proposed in [6] (an adaptive version of this strategy is presented in [5]) following [4] and [3]. [21] Y. If the frequencies of the disturbances are known. [16] . “Aircraft pitch control via second-order sliding technique. under the assumption of sinusoidal disturbances with known frequency. pp. a compensation scheme which allows for any initial frequency estimate greater than the given bound is presented in [17]. a global robust state feedback control scheme is presented for systems affected by both structured unknown disturbances and an unknown noise. pp. [6] T.1109/TAC. Rosier. Barbot. NOVEMBER 2005 relative degree by means of control continuous everywhere except this manifold. As far as nonlinear systems are concerned. 46. [12] S. and J. “Second order sliding mode control of a diesel engine. 2003.. Pan. regulators. 2002. Dordrecht. “Variable structure control with sliding sector. 924–941. Control. advances in variable structure systems. and W. Punta. on the other hand only the direct scheme guarantees exact disturbance compensation. when the only information on the controlled uncertain process is actually its relative degree. 2003. Jul. vol. 50. Pisano. the chattering effect is signiﬁcantly reduced. no. A.” Automatica. The real-time exact differentiator [16] of the appropriate order is combined with the proposed controller providing for the full SISO control based on the input measurements only.. Control..” Asian J. 211–228. Liapunov Functions and Stability in Control Theory. “Dynamic second-order sliding mode control of the hovercraft vessel. 1988. R.” Asian J. Ben-Asher. Control. “A survey of applications of second-order sliding mode control to mechanical systems. [5] A. output feedback. 12. semiglobal output feedback regulators have been described in [25]. J. pp. [17] . pp. I. [13] A. Control Optim. Bacciotti and L. As a result. 4. F. Floquet. “Sliding order and sliding accuracy in sliding mode control. and E. 2000. [3] G. I. no. no. Index Terms—Adaptive estimation. Gitizadeh. A. INTRODUCTION AND PROBLEM STATEMENT Over the last decades. REFERENCES [1] A. 5. no. Autom. pp. vol. Control Optim. and K. Z. for classes of NMP systems which are more general than those considered in this note. B. 10. 586–594. Kobayashi. Control. Shkolnikov. Las Vegas. S. phases. Autom. 39. 2nd ed. 36. Decision and Control. Technol.00 © 2005 IEEE . Control. 267.” AIAA J. Bhat and D. 498–504. pp. Santosuosso Abstract—A class of output feedback stabilizable nonlinear systems with known output dependent nonlinearities and affected by unknown sinusoidal disturbances is considered: Nonminimum phase systems are also allowed. 2079–2084.858647 0018-9286/$20. 47. 2005.” 00133 Rome. Hong. Decision and Control. V. magnitudes. 38. NO.” in Proc. 2003. The problem of designing a global output feedback compensator which drives the state of the system exponentially to zero is solved when the disturbance consists of a known number of biased sinusoids with any unknown bias. P. 1993.it. following earlier work in [20].-P.” SIAM J. Sliding Modes in Optimization and Control Problems.K.1816 IEEE TRANSACTIONS ON AUTOMATIC CONTROL. 11. Spurgeon. differentiation and output-feedback control. 58.” Int. “Singularly perturbed analysis of chattering in relay control systems. Scherpen.” in Proc.” Int. pp. 2000. M. Yaesh. pp. 1077–1083. A. no frequency considerations are relevant. Guid. Levantovsky. Djemai. several results are available under the minimum phase assumption (MP): The semiglobal output regulation problem is addressed in [24] for systems with unknown parameters in the exosystem. Nov. vol. 4. vol. and M. [10] A. Dec. Brown. no. Shtessel. “Adaptive VS differentiator. santosuosso@ing. K. 2001. pp. U. 1253–1271. “An asymptotic second-order smooth sliding mode control. 1247–1263. no. and frequencies. 2004. “Finite time stability of continuous autonomous systems. [8] K. Levant. pp. The simulation shows that it is probably the ﬁrst practically applicable output-feedback r-sliding controller with r > 2.” IEEE Trans. an early approach is based on the internal model principle [7]: The disturbances are viewed as outputs of known exosystems and may be rejected by incorporating the exosystem in the feedback path of the closed loop. Control. I. Pridor. 335–339. 9. pp. [14] A. 2002. 2002. HI. in [21]. Bartolini. vol. Sep. The global output tracking problem is studied in [2] for uncertain cascaded systems in lower triangular form coupled with a neutrally stable exosystem. Levant. If a positive lower bound for the frequency is known. 231–236. 23. 4. Manuscript received August 3. 2004. 2002. no. “Quasicontinuous high-order sliding mode controllers. vol. Recommended by Associate Editor J. [7] L.uniroma2. Khan. . nonlinear systems. 2003.” Int. “Smooth manifolds and high order sliding mode control. [15] A. Sarajevo. Furuta. Control Lett. Suzuki. Digital Object Identiﬁer 10. 4. K. 3. VOL. D. 9–12. revised October 5. pp. Fridman. Global Compensation of Unknown Sinusoidal Disturbances for a Class of Nonlinear Nonminimum Phase Systems Riccardo Marino and Giovanni L. Italy (e-mail: marino@ing. 7th VSS Workshop. [22] V.” Automatica.” IEEE Trans. vol. Filippov. Control. “Higher-order sliding mode stabilization for a class of nonholonomic perturbed systems. 2003. no. vol. 1992.” Au[18] tomatica. Furuta and Y. Preliminary results on the semiglobal regulation of nonminimum phase (NMP) systems are given in [26] and [12].it. vol. 1447–1451. 2000. vol. [2] J. Perruquetti. 46. In [18]. Bernstein. Levant (L. no. 43. B. vol. 43rd IEEE Conf. Sira-Ramírez. The Netherlands: Kluwer. no. Usai. [19] Y. New York: Springer-Verlag. “Universal SISO sliding-mode controllers with ﬁnite-time convergence. pp. The disturbance rejection in the case of a single sinusoidal disturbance with unknown frequency has been studied for stable linear systems in [1] and [17] and for minimum phase linear systems in [18] and [19]. 9/10. J. and S. 41st IEEE Conf. 76.

Sign up to vote on this title

UsefulNot useful- Quasi Continuous High Order Sliding Mode Controllersby Ulises Wally
- Discrete time sliding mode controlby Gholamreza Ilkhani
- LPV Model-based Robust Controller Design of Electroby Iwan Istanto
- Design Novel Lookup Table Changed Auto Tuning FSMC: Applied to Robot Manipulatorby AI Coordinator - CSC Journals

- Quasi Continuous High Order Sliding Mode Controllers
- Discrete time sliding mode control
- LPV Model-based Robust Controller Design of Electro
- Design Novel Lookup Table Changed Auto Tuning FSMC
- Speed control of Two Phase Induction Motor using Sliding Mode Controller
- 480-487
- Pet
- Characteristics of Effective Trainers
- Sliding Mode Based Integrated Guidance and Autopilot for Chasing UAV With the Concept of Time-Scaled Dynamic Inversion
- as2
- Final Course Schedule
- download_001.pdf
- Role Feedback
- Control Theory - Control-Intro PPT
- Bulling Towne
- 1.Intro
- Intro Control Systems
- Electrical Interview Questions-Interview Questions and Answers-25066
- 24134893
- Conceptual Statement
- phys-2007-1-4
- 360-degree-feedback.ppt
- Ch 6
- Strengths-Based Coaching Can Actually Weaken You
- How to Write Mathematics
- Recipient #102238 Printed on 1/4/2005. This is a ONE-TIME Copy.
- [Claude_Berge]_Topological_Spaces_Including_a_Tre(BookZZ.org).pdf
- LMI_PID-Robust PID Controller Design via LMI Approach-(Ge2002)
- Chiasso Nac March 98
- Performance Assessments That Motivate
- Carro

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue reading from where you left off, or restart the preview.