Accepted Manuscript Event-based reference tracking control of discrete-time nonlinear systems via delta operator method Rongjun Liu, Junfeng Wu, Dan Wang PII: DOI: Reference: S0016-0032(18)30500-3 https://doi.org/10.1016/j.jfranklin.2018.07.022 FI 3569 To appear in: Journal of the Franklin Institute Received date: Revised date: Accepted date: 16 November 2017 1 May 2018 31 July 2018 Please cite this article as: Rongjun Liu, Junfeng Wu, Dan Wang, Event-based reference tracking control of discrete-time nonlinear systems via delta operator method, Journal of the Franklin Institute (2018), doi: https://doi.org/10.1016/j.jfranklin.2018.07.022 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. ACCEPTED MANUSCRIPT Event-based reference tracking control of discrete-time nonlinear systems via delta operator method Rongjun Liua , Junfeng Wua,∗, Dan Wangb of Measure-Control Technology and Communication Engineering, Harbin University of Science and Technology, Harbin 150022 , P.R. China b College of Automation, Harbin Engineering University, Harbin 150001, P.R. China CR IP T a College Abstract AN US This paper investigates the event-based tracking control for delta-sampling systems with a reference model. Takagi-Sugeno (T-S) fuzzy model is used to approximate the nonlinearity. The delta operator is used to implement the discrete-time system. The event trigger is adopted for saving the network resources and the controller forces, and its detection period is designed with the same period of the delta-sampling period. Since the measurement is delayed from the sensor to the event-trigger, the methodology of time-delay systems, called the scaled small gain theorem, is applied for the system stability analysis. The reference output tracking controller is designed to ensure the stability of the resulting system in H∞ sense. The optimization conditions of the desired H∞ event-based tracking controller are synthesized, and the simulation example validates its effectiveness finally. M Keywords: Event-triggered control, output tracking control, fuzzy systems, delta operator. 1. Introduction AC CE PT ED Tacking control has been one of the hot spots in control engineering field. In the past decades, the tracking control problems of linear/nonilear control systems, uncertain systems, switching systems, and networked control systems, have been well studied [1, 2, 3, 4, 5]. The aim of output tracking control is to design a controller to force the plant to track a output signal of a given reference model. Generally, the real plants are usually complex with nonlinearity and uncertainties including the parameter perturbation and disturbance noise. Therefore, the modeling of the plants are primarily important for accurate sensor sampling and controller design. As is known, Takagi-Sugeno (T-S) fuzzy models [6] have been fully applied for the representation of nonlinear systems since the convenience of model approximation and controller realization [7]. In the frame of the T-S fuzzy model, several fuzzy rules based on the fuzzy sets are built with corresponding linear subsystems. In this way, nonlinearity and uncertainties can be approximated in pieces, and then a defuzzified plant can be yielded. The application of the T-S fuzzy model can be found for different kind of the control systems such as the static output feedback control [8], state feedback control [9, 10], observer-based control[11, 12, 13]. In digital control systems and the communications fields, high-speed signal processing can store informative data and deal with the fast signal transmission [14, 15, 16]. The analysis methods based on traditional shift operator can handle the intractable digital control problems when fast sampling frequency are executed [17, 18]. To formulate the systems of the continuous-time form and the discrete-time form in a unit frame, delta operator was detailed in [17, 19]. With a fast sampling period based on delta operator, the finite word ∗ Corresponding author Email addresses: 641445222@qq.com (Rongjun Liu), wu_jf@hrbust.edu.cn (Junfeng Wu), bigbigfaceshop@163.com (Dan Wang) Preprint submitted to Journal of The Franklin Institute August 13, 2018 ACCEPTED MANUSCRIPT length performance can operate well. The following definition gives the formulation of delta operator [17]: ( d x(t ), h = 0, dtk k δx(tk ) = (1) x(tk +h)−x(tk ) , h 6= 0, h AC CE PT ED M AN US CR IP T where tk = kh with h the sampling period, and δx (tk ) is the delta operator of x(tk ). x(tk ) is the state variable of a real system. Obviously, with a small sampling period, the control system closes to a continuous-time system. Thus, the characteristics of the stability and performance can be well responded like the continuoustime systems. By utility of the delta operator, typical control methods, such as feedback control [20, 21], fault-tolerant control [22], disturbance rejection control [23] and tracking control [24], have designed for the linear and nonlinear systems. For instance, in terms of the discrete-time fuzzy systems, the authors in [25] coped with the H∞ control problem based on delta operator. The work in [26] presented a saturated control scheme by using the fuzzy models and delta operator. For the discrete-time nonlinear systems based on delta operator, an output tracking control scheme was detailed in [24] using T-S fuzzy models. Recently, in contrast of sending information with a fixed period, the event-triggering communication strategy has been built under the consideration of saving energy economy and reducing transmissions task in the network resource [27, 28]. Frequent signal transmissions can deplete sensor or controller batteries more easily, especially in the case when networks are utilizing the periodic communication paradigm. The general setting of an event-triggering mechanism consists of an event detector unit and a sender unit. The detector unit monitors a user-defined event-triggering condition, and instructs the sender unit to transmit new measurements on violation of the condition. It deserves to be mentioned that the authors in [29] discussed the transmission of a a discrete event-triggered communication of fuzzy-model-based systems. In [30], an event-driven control strategy was presented for the consensus of multi-agent systems with fixed/switching topologies. The work in [31] introduced a methodology in terms of a class of time-delay system to handle the event-triggered control of networked control systems. For the multi-area power systems with communication delays, load frequency control was detailed in [32] based on an event-triggering condition. In terms of the sampling systems based on delta operator, however few techniques are developed to the event-triggering mechanism. Meanwhile, event-based tracking control of the nonlinear system still reminds an open problem to explore. Motivated from the above discussion, we attempt to explore an event-based tracking controller for the nonlinear systems of delta-sampling form. In this work, we focus on dealing with the problems of the event-triggering mechanism design and the full state-feedback tracking controller synthesis. The contribution follows that: i) the event-triggering mechanism based on delta-sampling form is presented with the event-detecting period being concordant with the delta-operator sampling period, ii) the stability criterion based on scaled small gain theorem is developed for the resulting event-driven control system, and iii) an optimal event-based output tracking controller is designed for the desired H∞ performance with a predefined reference system. The rest of the paper follows that Section 2 explain the system descriptions, Section 3 presents the mentioned control methods, and Section 4 gives a illustrative example for validation. The conclusion is detailed in Section 5. Notation: `2 [0, ∞) is the space of square-integrable vector functions over [0, ∞). ∆1 ◦ ∆2 denotes the series connection of the mappings ∆1 and ∆2 . For matrix X ∈ Rn×n , [X]s is used to denote X + X T for simplicity. The superscripts “X T ” and “X −1 ” are respectively the matrix transpose and inverse of X. The transposed elements in the matrix symmetric positions are denoted by “?”. “I” is an appropriately dimensioned identity matrix if not specified. 2. Problem Formulation Based on the formulation of delta operator (1), we establish the following T-S fuzzy systems Plant Rule i : IF g1 (tk ) is Si1 , and · · · , and gp (tk ) is Sip , THEN δx(tk ) = Ai x(tk ) + Bi u(tk ) + Bwi w(tk ), y(tk ) = Ci x(tk ) + Di u(tk ) + Dwi w(tk ), 2 (2) ACCEPTED MANUSCRIPT CR IP T where x(tk ) ∈ Rn and u(tk ) ∈ Rm are the system state and system control input, y(tk ) ∈ Rm and w(tk ) ∈ Rl are the system controllable output and the bounded external disturbance input. Matrices Ai , Bi , Bwi , Ci , Di and Dwi are known. Sij and gj (tk ) are respectively the fuzzy set and the premise variable (i = 1, 2, · · · , q, j = 1, 2, · · · , p). The defuzzified expression of model (2) gives q P δx(tk ) = vi (g(tk )) [Ai x(tk ) + Bi u(tk ) + Bwi w(tk )] , i=1 q P y(tk ) = vi (g(tk )) [Ci x(tk ) + Di u(tk ) + Dwi w(tk )] , (3) i=1 q P µi (g(tk )), vi (g(tk )) = µi (g(tk ))/ i=1 where µi (g(tk )) = p Q Sij (gj (tk )), j=1 q P i=1 vi (g(tk )) = 1, and 0 ≤ µi (g(tk )) ≤ 1 is the membership function T AN US corresponding to i-th rule for fuzzy set Sij . In the event trigger unit, the event detector monitors the state x(tk ) with the sampling period h. Let the current detected state as x(ti ) (i = 0, 1, 2, . . . , ∞), and the latest detected and triggered one as x(tik ) (ik ∈ N, i0 = 0, k ∈ N is the number of triggering). The following event trigger condition (4) is adopted to trigger new measurement over the network once x(tik +j ) satisfies (x(tik +j ) − x(tik )) (x(tik +j ) − x(tik )) > ρxT (tik )x(tik ), (4) M where ρ is a user-defined error tolerance. Assume that the initial state x(t0 ) is transmitted (known), and then the next triggered time instant tik+1 is determined by tik+1 = tik + max dk , h , (5) where n o T dk = min tj | (x(tik +j ) − x(tik )) (x(tik +j ) − x(tik )) > ρxT (tik )x(tik ) . ED j≥1 PT Remark 1. Actually, Zeno phenomena [33, 34, 35] are a common problem in event-triggering mechanism for the control loop. Motivated by the seting in [33], we modify the positive real number as the sampling period h in (5) to refrain from the Zeno behavior in the presented event-triggered control loop, since the minimum step of the triggering interval is the sampling period via the delta operator in the control loop. CE Let sk is the time instant when the measurement x(sk ) arrives at the ZOH. As a result, on time interval [sk , sk+1 ) for k ∈ N+ , we obtain that the following inequality always holds for j = 1, 2, · · · T (x(tik +j ) − x(tik )) (x(tik +j ) − x(tik )) ≤ ρxT (tik )x(tik ). (6) AC Considering the delayed state induced by the event-triggering mechanism, let dk = tik −tk be the delayed time of k-th triggered measurement between the event trigger and the ZOH. Assume dm = min{dk |k = 0, 1, 2, . . . , ∞} and dM = max{dk |k = 0, 1, 2, . . . , ∞} be the minimum and the maximum delay, respectively. On interval [sk , sk+1 ) for k ∈ N+ , define the following sub-intervals: I1 = [sk , sk + h), I2 = [sk + h, sk + 2h), .. . ? Ijk = [sk + (jk? − 1)h, sk+1 ), 3 ACCEPTED MANUSCRIPT with jk? = min+ {j | sk + (j − 1)h ≥ sk+1 }, and functions j∈N CR IP T tk − tik , tk ∈ I1 , tk − tik +1 , tk ∈ I2 , τk = .. . tk − tik +(jk? −1) , tk ∈ Ijk? , x(tik ) − x(tik ), tk ∈ I1 , x(tik ) − x(tik +1 ), tk ∈ I2 , e(tk ) = .. . x(tik ) − x(tik +(jk? −1) ), tk ∈ Ijk? . (7) (8) Then from (7), we know τk is a piecewise linear function, and the interval [sk , sk+1 ) can be represented as ? [sk , sk+1 ) := jk [ Ij . j=0 AN US Considering the measurement transmitted through the network each time, one can obtain that it always holds from (7) that dm ≤ τk < dk + h ≤ dM + h. Moreover, from (7) and (8) we know that the input signal x̄(tk ) of the controller is x̄(tk ) = x(tik ) = e(tk ) + x(tk − τk ), tk ∈ [sk , sk+1 ). And then the condition (6) can be updated as T (9) M eT (tk )e(tk ) ≤ ρ (e(tk ) + x(tk − τk )) (e(tk ) + x(tk − τk )) ED for any tk ∈ [sk , sk+1 ). Consequently, the received signal of the ZOH equals to the triggered the signal at event-triggering instants t0 < ti1 < ti2 < · · · < tik < · · · . In this paper, we consider the reference model and the reference output yr (tk ) ∈ Rm as follows δxr (tk ) = Exr (tk ) + r(tk ), (10) yr (tk ) = F xr (tk ), CE PT where xr (tk ) and r(tk ) are respectively the reference state and the reference input which is bounded. E and F are constant matrices with E Hurwitz. Based on the triggered measurement x(tik ), using the parallel distributed compensation, we design the following event-based tracking controller to force the system (2) to track the reference output (see the reference in (10)). AC u(tk ) = = q X i=1 q X i=1 vi (g(tk )) [Ki x(tik ) + Kri xr (tk )] (11) vi (g(tk )) [Ki (e(tk ) + x(tk − τk )) + Kri xr (tk )] , (12) where tk ∈ [sk , sk+1 ), matrices Ki and Kri are the tracking controller parameters. T Let the tracking error ỹ(tk ) = y(tk )−yr (tk ) and state z(tk ) = xT (tk ) xT . Considering (3)-(11), r (tk ) for any tk ∈ [sk , sk+1 ) the closed-loop system gives q P q P δz(tk ) = vi vj Āij z(tk ) + B̄dij z (tk − τk ) + B̄ij $(tk ) , i=1 j=1 (13) q P q P vi vj C̄ij z(tk ) + D̄dij z (tk − τk ) + D̄ij $(tk ) , ỹ(tk ) = i=1 j=1 4 ACCEPTED MANUSCRIPT T and where vi = vi (g(tk )) for simplicity, $(tk ) = eT (tk ) wT (tk ) rT (tk ) Ai Bi Krj Bi Kj 0 Bi Kj Bwi 0 Āij = , B̄dij = , B̄ij = , 0 E 0 0 0 0 I C̄ij = Ci + Di Krj − F , D̄dij = DKj 0 , D̄ij = DKj Dwi 0 . Before further we introduce the following definition and lemmas. CR IP T Definition 1. [36] The stability of a delta operator system can be ensured when the following condition holds: 1. V (x(tk )) ≥ 0, with equality if and only if x(tk ) = 0; 2. δV (x(tk )) = [V (x(tk + h)) − V (x(tk ))]/h < 0, where V (x(tk )) is a Lyapunov function in δ-domain. The first condition gives the stability criterion in terms of the Lyapunov functions both in s-domain and z-domain. With respect to the condition 2, there exists dV (x(tk )) V (x (tk + h)) − V (x(tk )) = < 0, h→0 h dtk lim δV (x(tk )) = lim h→0 AN US when h → 0. It corresponds to the Lyapunov function in z-domain. When h = 1, there exists δV (x(tk )) = [V (x(tk + 1)) − V (x(tk ))]/1 = V (x(tk + 1)) − V (x(tk )) < 0. Hence, the Lyapunov function in δ-domain implies traditional z-domain when h = 1. Lemma 1. [36] For any of the vector functions a (tk ) and b(tk ) it holds that δ a(tk )b(tk ) = δa(tk ) · b(tk ) + a(tk ) · δb(tk ) + h · δa(tk ) · δb(tk ). Lemma 2. For any matrix Z > 0 and any positive integers q and q0 with 1 ≤ q0 ≤ q, it holds that xT (l) X l=q0 q X M q X l=q0 x (l) ≤ (q − q0 + 1) q X xT (l) Xx (l) . l=q0 ED Lemma 3. (Scaled small gain theorem) [37, 38, 39] Consider the following interconnection system β(tk ) = Γ · α(tk ), (14) α(tk ) = ∆ · β(tk ), (15) CE PT where operator Γ mapping α(tk ) ∈ Rα to β(tk ) ∈ Rβ . Subsystem (14) is a known system and the subsystem (15) is an unknown one but the operator ∆ ∈ = , k∆k∞ ≤ 1. Given that subsystem (14) is internally stable, the interconnection system (14)–(15) is asymptotically stable for all ∆ ∈ = if Nβ ◦ Γ ◦ Nα−1 ∞ < 1, where nonsingular matrices {Nα , Nβ } ∈ X, and X , {Nα , Nβ } ∈ Rα×α · Rβ×β : kNα ◦ Γ ◦ Nβ−1 k∞ < 1 . To represent the time-delay system to an interconnection system [40], we define AC α(tk ) = dM 1 [2z (tk − τk ) − (z(tk − dm ) + z(tk − dM − h))] + h − dm to approximate z (tk − τk ). Further, letting d = dM + h − dm , it results that 1 2 α(tk ) = z (tk − τk ) − [z(tk − dm ) + z (tk − dM − h)] d 2 −τk /h−1 −dm /h−1 X X 2 h h = δz (tk + lh) − δz (tk + lh) d 2 2 l=−dM /h−1 = 1 dM /h − dm /h i=−τk /h −dm /h−1 X φ (l) δz (tk + lh) , l=−dM /h 5 (16) ACCEPTED MANUSCRIPT where α(tk ) is the estimating error, and φ (l) = 1, (l ≤ τk /h) , −1, (l > τk /h) . According to Lemma 3, we write the system (13) as δz(tk ) z̃(tk ) β(tk ) = Γ(tk ) · α(tk ) , ỹ(tk ) γ$(tk ) z T (tk ) z T (tk − dm ) z T (tk − dM − h) Ā(tk ) = C̄(tk ) = B̄(tk ) = q X q X i=1 j=1 q X q X i=1 j=1 q q X X , tk = −dM /h, −dM /h + 1, · · · , 0, and 1 d 2 B̄d (tk ) 2 B̄d (tk ) 1 d B̄ (t ) d k 2 2 B̄d (tk ) 1 d D̄ (t ) d k 2 2 D̄d (tk ) q q XX 1 2 B̄d (tk ) 1 2 B̄d (tk ) 1 2 D̄d (tk ) Ā(tk ) Γ(tk ) = Ā(tk ) C̄(tk ) T CR IP T (18) γ −1 B̄(tk ) γ −1 B̄(tk ) , γ −1 D̄(tk ) AN US where z̃(tk ) = α(tk ) = ∆(tk ) · β(tk ), (17) vi vj B̄dij , vi vj Āij , B̄d (tk ) = vi vj C̄ij , D̄d (tk ) = vi vj B̄ij , D̄(tk ) = vi vj D̄di , i=1 j=1 q X q X vi vj D̄ij . i=1 j=1 M i=1 j=1 i=1 j=1 q X q X ED P−dm /h−1 1 and it can be specified the mapping ∆(tk ) : β(tk ) → α(tk ) = dM /h+1−d i=−dM /h+1 φ (i) β (tk + ih) m /h according to the mapping in (15). Referring to [40, 41], we know operator k∆(tk )k∞ ≤ 1. Meanwhile, we present the following lemma for stability analysis basis. PT Lemma 4. Suppose system (17) is internally stable. The closed-loop system (14) is asymptotically stable in for the mapping ∆(tk ), if there exists strictly positive-definite {Nα , Nβ } ∈ X such that an H∞ sense Nβ ◦ Γ ◦ Nα−1 < 1. ∞ CE Proof. In terms of the interconnection system (17)–(18), it results form Lemma 3 that 2 2 2 2 kβ(tk )k2 + kỹ(tk )k2 < kα(tk )k2 + kγ$(tk )k2 . (19) AC P−dm /h−1 1 Since α(tk ) = dM /h+1−d i=−dM /h+1 φ (i) δz (tk + ih), it derives kỹk2 ≤ γ k$(tk )k2 ∀ 0 6= $ ∈ `2 [0, ∞), m /h which means interconnection system (17) with an H∞ performance γ for ∆(tk ) ∈ X. This ends the proof. Therefore, in the next step we will analyze the stability of the system (17), i.e. the stability of the following system δz(tk ) = Ā(tk )z(tk ) + 21 B̄d (tk ) [z(tk − dm ) + z(tk − dM − h) + dα(tk )] + B̄(tk )γ$(tk ), (20) ỹ(tk ) = C̄(tk )z(tk ) + 21 D̄d (v) [z(tk − dm ) + z(tk − dM − h) + dα(tk )] + D̄(tk )γ$(tk ). Hence, the aim of this work is to analyze the stability of the resulting augmented system (20), and then design the parameters of the event trigger and the tracking controller. Additionally, the term z (tk − τk ) in (13) is actually a time-varying delay term. In this paper we apply the so called analysis methodology in terms of the “ time-delay system” to analyze and synthesize the resulting event-based control system (20). 6 ACCEPTED MANUSCRIPT 3. Main Results Firstly, based on the resulting time-delay control system (20), we present the following theorem for a stability criterion using Lyapunov functional approach. CR IP T Theorem 1. System (17)–(18) is asymptotically stable in an H∞ sense, if there exist symmetric matrices P > 0, Y > 0, U1 > 0, U2 > 0, W1 > 0, W2 > 0 and Z > 0 such that Φ1 Φ2 Υ= < 0, (21) ? Φ3 where Φ1 = −Z ? 0 −I , Φ2 = 0 0 X Ā(tk ) C̄(tk ) 1 X B̄d (tk ) 2 1 D̄ (t ) 2 d k 1 X B̄d (tk ) 2 1 D̄ (t ) 2 d k Ω1 = (h − 2)P + dm W1 + (dM + h)W2 , d X B̄d (tk ) 2 d D̄ (t ) 2 d k X B̄(tk ) D̄(tk ) AN US 1 1 W1 − W2 , Ω2 = [P Ā(tk )]s + (dM /h + 1 − dm /h + 1)Y + U1 + U2 − dm dM + h 1 1 d Ω1 P Ā(tk ) P B̄d (tk ) P B̄d (tk ) P B̄d (tk ) 2 2 2 1 1 1 d 1 ? P B̄d (tk ) + dm W1 P B̄d (tk ) + dM +h W2 P B̄d (tk ) Ω2 2 2 2 ? ? − 41 Y − U1 − d1m W1 − 14 Y − d4 Y Φ3 = 1 1 − d4 Y ? ? − 4 Y − U2 − dM +h W2 ? 2 d ? ? ? ? − 4 Y −Z ? ? ? ? ? Proof. Select Lyapunov-Krasovskii functional candidate V (tk ) = 4 P P B̄(tk ) P B̄(tk ) 0 0 0 −γ 2 I , . Vl (tk ) for system (20), where V1 (tk ) = z T (tk )P z(tk ), dM /h+1 i=dm /h j=1 dm /h V3 (tk ) = h X dM /h+1 z T (tk − ih)U1 z(tk − ih) + h PT i=1 z T (tk − jh)Y z(tk − jh), ED i X X V2 (tk ) = h M l=1 dm /h V4 (tk ) = h2 X i X i=1 δz(tk − jh)W1 δz(tk − jh) + h2 CE i=1 j=1 X z T (tk − ih)U2 z (tk − ih) , dM /h+1 i X X i=1 j=1 β(tk − jh)W2 β(tk − jh). Using Lemma 1 with system (20), δV1 (tk ) results that AC δV1 (tk ) = 2δ T (z(tk ))P z(tk ) + hδ T (z(tk )) P δ(z(tk )) n = hδ T (z(tk ))P δ (z(tk )) + 2z T (tk )P Ā(tk )z(tk ) o 1 + B̄d (tk ) [z(tk − dm ) + z(tk − dM − h) + dα(tk )] + B̄(tk )$(tk ) . 2 7 (22) ACCEPTED MANUSCRIPT By the delta operator manipulation, it follows that δV2 (tk ) ≤ (dM /h + 1 − dm /h + 1) z T (tk )Y z(tk ) − z T (tk − αT ) Y z (tk − αT ) = (dM /h + 1 − dm /h + 1) z T (tk )Y z(tk ) T 1 d 1 − z(tk − dm ) + z (tk − dM − h) + α(tk ) 2 2 2 1 1 d ×Y z(tk − dm ) + z(tk − dM − h) + α(tk ) , 2 2 2 dm /h X 1 z T (tk − (i − 1) h) U1 z (tk − (i − 1) h) δV3 (tk ) = · h h i=1 X i=1 z T (tk − (i − 1) h) U2 z (tk − (i − 1) h) dm /h i=1 dM /h+1 z T (tk − ih)U1 z (tk − ih) + X T T i=1 z T (tk − ih) U2 z(tk − ih) AN US − X CR IP T dM /h+1 + (23) = z (tk ) (U1 + U2 ) z(tk ) − z (tk − dm )U1 z(tk − dm ) − z T (tk − dM − h)U2 z (tk − dM − h) . Using Lemma 2, δV4 (tk ) results that (24) δV4 (tk ) ≤ β T (tk ) ((dm /h)hW1 + (dM /h + 1) hW2 ) β(tk ) dm /h dm /h X h X T β (tk − ih) W1 β(tk − ih) dm /h i=1 i=1 − h dM /h + 1 M − dM /h+1 β T (tk − ih) W2 ED X i=1 dM /h+1 X i=1 β (tk − ih) PT = δ T (z(tk )) (dm W1 + dM W2 ) δ(z(tk )) 1 T − [z(tk ) − z(tk − dm )] W1 [z(tk ) − z(tk − dm )] dm 1 T [z(tk ) − z(tk − dM − h)] W2 [z(tk ) − z (tk − dM − h)] . − dM + h CE (25) AC For any P > 0, it always holds that n 0 = −2δ T (z(tk ))P δz(tk ) − Ā(tk )z(tk ) o 1 − B̄d (tk ) [z(tk − dm ) + z(tk − dM − h) + dα(tk )] − B̄(tk )$(tk ) . 2 (26) Further, recalling Lemma 4, we consider the following inequality from (19) for H∞ performance analysis. z= ∞ X T β (tk ) Xβ(tk ) − αT (tk )Xα(tk ) + ỹ T (tk )ỹ(tk ) − γ 2 $T (tk )$(tk ) . (27) k=0 With ϑ = δ T (z(tk )) z T (tk ) z T (tk − dm ) z T (tk − dM − h) αT (tk ) γ$T (tk ) 8 T , under the zero ACCEPTED MANUSCRIPT initial condition, it results that z ≤ z + V (∞) − V (0) ∞ X T = β (tk )Xβ(tk ) − αT (tk )Xα(tk ) + ỹ T (tk )ỹ(tk ) − γ 2 $T (tk )$(tk ) + δV (tk ) = k=0 ∞ X k=0 −1 ϑT (tk )(Φ3 − ΦT 2 Φ1 Φ2 )ϑ(tk ). CR IP T −1 According to Schur complement, Υ < 0 is equivalent to Φ3 − ΦT 2 Φ1 Φ2 < 0. Hence z < 0 under the zero initial condition. With Lemma 4 we know that kỹ(tk )k2 < kγ$(tk )k2 holds and the H∞ performance index γ guaranteed. Additionally, considering α(tk ) γ$(tk ) = 0, it follows δV (tk ) < 0 for any nonzero states. That is to say the asymptotic stability of system (20) can be ensured. It ends the proof. In terms of the conditions (21), for a more relaxed stability condition of system (20), the following corollary is developed. AN US Corollary 1. System (17)–(18) is asymptotically stable in an H∞ sense, if there exist proper symmetric matrices P > 0, Y > 0, U1 > 0, U2 > 0, W1 > 0, W2 > 0 and Z > 0, for i, j = 1, 2, · · · , q such that: Υii < 0, (28) 1 1 Υii + (Υij + Υji ) < 0, (i 6= j) , q−1 2 Υij ? ? ? ? ? 1 P B̄di 2 1 1 P B̄ + W2 di 2 dM +h 1 −4Y − 41 Y − U2 − dM1+h W2 ? ? d P B̄di 2 d P B̄di 2 − d4 Y − d4 Y 2 − d4 Y − Z ? P B̄i P B̄i 0 0 0 −γ 2 I X B̄i , D̄i 1 1 W1 − W2 . = P Āij + ĀT ij P + (dM /h + 1 − dm /h + 1) Y + U1 + U2 − dm dM + h Φ2ij Φ3ij , Φ2ij = 0 0 X Āij C̄ij 1 X B̄di 2 1 D̄ 2 di 1 X B̄di 2 1 D̄ 2 di d X B̄di 2 d D̄ 2 di PT Ω2ij = ? ? Φ1 = ? 1 P B̄di 2 1 P B̄ + d1m W1 di 2 1 − 4 Y − U1 − d1m W1 P Āij Ω2ij ? ? ED Φ3ij Ω1 ? ? ? M where (29) , Proof. Firstly, based on (28)–(29) we know Υij < 0 holds for i, j = 1, 2, · · · , q, using the results in [42]. q P q P Moreover, substituting the defined matrices of system (20) into Υij yields vi vj Υij = Υ. Hence, Υ < 0 i=1 j=1 CE holds, and the same conclusion in Theorem 1 can be obtained. This ends the proof. AC Remark 2. The conditions synthesized in Theorem 1 and Corollary 1 are actually with some couplings of the tracking controller gains Ki and Kri . To find some solutions to the controller gains, we develop the conditions in Corollary 1 by matrix transformation technique. Theorem 2. System (17)–(18) is asymptotically stable with an H∞ performance level γ, if there exist appropriately dimensioned symmetric matrices Q > 0, R > 0, Ȳ > 0, Ū1 > 0, Ū2 > 0, W̄1 > 0, W̄2 > 0, Z̄ > 0, Mj and Lj for i, j = 1, 2, · · · , q the conditions (31) and (32) hold. The H∞ performance level γ can be optimized by min ρ = γ 2 subject to (30) Ῡii < 0, 1 1 Ῡii + Ῡij + Ῡji < 0, (i 6= j) q−1 2 9 (31) (32) ACCEPTED MANUSCRIPT where Φ̄2ij = Φ̄3ij with = Φ̄1 ? Φ̄2ij Φ̄3ij 0 ΘAij 0 ΘCij Ω̄1 ? ? ? ? ? , Φ̄1 = Z̄ − 2Θ0 ? 0 −I 1 d 2 ΘBdi 2 ΘBdi 1 d Θ Ddi 2 2 ΘDdi 1 2 ΘBdi 1 1 Θ 2 Bdi + dm W1 − 41 Ȳ − Ū1 − d1m W̄1 1 2 ΘBdi 1 2 ΘDdi ΘAij Ω̄2ij ? ? ? ? , ΘBi ΘDi , 1 2 ΘBdi 1 1 2 ΘBdi + dM +h W2 − 14 Ȳ 1 − 4 Ȳ − Ū2 − dM1+h W̄2 ? ? ? d 2 ΘBdi d 2 ΘBdi − d4 Ȳ − d4 Ȳ d2 − 4 Ȳ − Z̄ ΘBi ΘBi 0 0 0 −γ 2 I , CR IP T Ῡij = ? ? ? U11 U12 U21 U22 Ȳ = , Ū1 = , Ū2 = , ΘDdi = Ddi Q 0 , ? U13 ? U23 W11 W12 W21 W22 X1 X2 W̄1 = , W̄2 = , Z̄ = , ? W13 ? W23 ? X3 Ai Q + Bi Mj Bi Nj Bdi Q 0 Q 0 ΘAij = , ΘBdi = , Θ0 = , 0 ER 0 0 0 R Bdi Mj Bwi 0 ΘBij = , ΘCij = Ci Q + Di Mj Di Nj − F R , 0 0 I ΘDij = Bdi Mj Dwi 0 , Ω̄1 = (h − 2)Θ0 + dm W̄1 + dM + hW̄2 , 1 1 W̄1 − W̄2 . Ω̄2ij = ΘAij + ΘTAij + (dM /h + 1 − dm /h + 1) Ȳ + Ū1 + Ū2 − dm dM + h Y2 Y3 M AN US Y1 ? ED If the conditions (31) and (32) hold, then the controller gains in (11) are determined by Kj = Mj Q−1 , Krj = Lj R−1 . (33) CE PT Proof. Following the proof of Corollary 1, for i, j = 1, 2, · · · , q we have Ῡij < 0 from (31) and (32). Let −1 P̄ −T Z̄ P̄ −1 , P̄ = P −1 = diag {Q, R}, and define Y = P̄ −T Ȳ P̄ −1 , U1 = P̄ −T Ū1 P̄ −1 , U2 = P̄ −T Ū 2 P̄−1 , Z = −1 −T −1 −1 −T W1 = P̄ W̄ , W2 = P̄ W̄2 P̄ . Then by the congruence transformation diag P̄ , I, P̄ , P̄ −1 , P̄ −1 , 1 P̄ P̄ −1 , P̄ −1 , I to Ῡij < 0, we have Φ̃1 Φ̃2ij < 0, (34) ? Φ3ij AC where Mj = Kj Q and Lj = Krj R from (33), and X − 2P 0 0 P Āij Φ̃1 = , Φ̃2ij = ? −I 0 C̄ij 1 2 P B̄dij 1 2 D̄dij 1 2 P B̄dij 1 2 D̄dij d 2 P B̄dij d 2 D̄dij P B̄ij D̄ij . T Besides, for any X > 0, since (X − P ) X −1 (X − P ) ≥ 0 it results −P X −1 P ≤ X − 2P . Using this property, it yieldsΥij < 0 in Corollary 1, by performing the congruence transformation via diag XP −1 , I, I, I, I, I, I, I . The proof is completed. Remark 3. Compared with the existing work [24, 43, 2], the proposed controller is an event-triggered one for the output tracking control. It can save the control cost and communication resource since its the communication between the controller the plant is aperiodic. The simulations in Section 4 will give the illustration of this virtue. Besides, the controller gains designed in this paper can be changed (subject to the conditions synthesized in Theorem 2) when differen sampling periods of the plant are used in practice. 10 ACCEPTED MANUSCRIPT Remark 4. Since solution of the synthesised conditions in Theorem 2 is subject to the sampling period h, different sampling periods may yield different optimal γ. It depends on the specific circumstances for a real control system. In simulations, we will use a fixed sampling period for a clear comparison between the presented event-based controller and the conventional time-driven controller. 4. Simulation Example CR IP T We provide the mass-spring mechanical system [43] as the plant for the demonstration of the proposed event-based tracking controller. Let s denote the displacement from a reference point. According to the dynamics equation ms̈ + Ff + Fs = u, where Ff = cṡ with c > 0 stands for the viscous damping force, Fs = k 1 + a2 s2 s with constants k and a stands for the restoring force of the spring, m stands for the mass, and u stands for the external control input. Thus, the dynamics equation is AN US ms̈ + cẋ + k ṡ + ka2 s3 = u, T T . The state-space equation yields = s ṡ Let the state variable vector x = x1 x2 ẋ1 0 1 x1 0 = + u. ẋ2 −k − ka2 x21 −c x2 1 The system output is give by y = x1 . Considering the noise in the state process, we apply the following fuzzy systems to approximate the plant with h = 0.1 s. Fuzzy Rule 1 : IF x21 (tk ) is D̄, THEN δx(tk ) = A1 x(tk ) + B1 u(tk ) + Bw1 w(tk ), y(tk ) = C1 x(tk ) + D1 u(tk ) + Dw1 w(tk ). M ¯ THEN Fuzzy Rule 2 : IF x21 (tk ) is d, δx(tk ) = A2 x(tk ) + B2 u(tk ) + Bw2 w(tk ), y(tk ) = C2 x(tk ) + D2 u(tk ) + Dw2 w(tk ). ED The membership functions are chosen as v1 x21 (tk ) = x21 − d¯ / D̄ − d¯ , v2 x21 (tk ) = 1 − v1 x21 (tk ) , AC CE PT where D̄ = 0.5 and d¯ = 0 are the upper bound and lower bound of x21 , respectively. With c = 0.4 kg/s, a2 = 0.9, m = 1 kg, k = 1.1 kg/s2 , we have −0.0631 0.9825 0.0395 0.4 A1 = , B1 = , Bw1 = , −1.5671 −0.4561 0.9825 0.2 −0.0435 0.9830 0.0396 0.6 A2 = , B2 = , Bw2 = , −1.0813 −0.4367 0.9830 0.3 T C1 = 1 0 , D1 = 0.1, Dw1 = 0.01, C2 = C1 , D2 = D1 , Dw2 = Dw1 . The reference model is chosen as δxr (tk ) = −0.9516xr (tk ) + r(tk ), yr (tk ) = 0.15xr (tk ). Suppose dm = 0.2 s and dM = 2.0 s and set the event trigger parameter ρ = 1.5. Solving the optimization problem in Theorem 2, we have γmin = 0.7381, and the presented event-triggered controller: 2 X u(tk ) = vi Ki x(tik ) + Kri xr (tk ) , i=1 11 ACCEPTED MANUSCRIPT with the gains K1 = K2 = −1.8641 −2.6672 −1.2257 −1.1334 , Kr1 = 0.3421, , Kr3 = 4.2457. AN US CR IP T To clear illustrate the efficacy of the designed event-triggered controller, we compare the presented controller with the conventional controller in [24]. Meanwhile, we use the same setting and system parameters of the simulated systems for the both controllers. Assumed that w(tk ) = 0.1rand · sin (20tk ), r(tk ) = 0.2 sin (20tk ). Then the simulation results are illustrated in Figs. 1–4. The control inputs used in the two cases are given in Fig. 1. Fig. 2 shows the event-triggered intervals at each triggered instant of the event-triggered controller. From Fig. 2 we know there are 26 times (including the one at the initial instant) of triggering in the event trigger unit. That is to say it saves almost (20/h − 26)/(20/h) ≈ 98% cost for controller updating. Fig. 3 shows that y(tk ) tracks yq (tk ) well under the conventional controller in [24] (shown in Fig. 3(a)) and the designed event-triggered controller (shown in Fig. 3(b)). Accordingly, Fig. 4 depicts both tracking errors ỹ. The tracking errors are almost the same. Obviously, the tracking performance under the presented controller is almost as well as the conventional one while it saves much control cost. 1 0.5 M −0.5 −1 −1.5 ED C ontroller signa l u 0 PT −2 T he conventio nal contro ller in [2 4 ] −2.5 0 5 10 tk 15 20 Figure 1: Control input u(tk ). AC CE −3 T he designed event-triggered co ntro ller 5. Conclusion The problem of the event-based tracking control of discrete-time nonlinear systems has been handled in this article. The delta-sampling plant based on T-S fuzzy model has been formulated. The event-triggering mechanism has been designed with the same period of the delta-sampling period. For the resulting timedelay system, Lyapunov stability theory and the scaled small gain theorem have been based for the stability analysis. The reference output tracking controller has been designed to ensure the stability of the resulting system in H∞ sense. Finally, the designed event-based control has been successfully validated by the massspring mechanical tracking control system. Based on the developed results, future work will be concerned 12 ACCEPTED MANUSCRIPT 2 1.8 1.6 CR IP T 1.4 τk 1.2 1 0.8 0.6 AN US 0.4 0.2 0 0 5 10 tk 15 20 (a) 0.8 M Figure 2: Triggered intervals of the event trigger at each triggered instant. ED y(tk ) under the co ntroller in [24 ] yr (tk ) under the co ntroller in [24 ] 0.7 y(tk ) under the desig ned co ntro ller yr (tk ) under the desig ned co ntro ller 0.7 0.6 PT 0.6 (b) 0.8 0.5 0.5 0.4 0.4 CE 0.3 0.3 0.2 AC 0.2 0.1 0.1 0 0 −0.1 −0.1 0 5 10 tk 15 20 −0.2 0 5 Figure 3: Outputs y(tk ) and yr (tk ). 13 10 tk 15 20 ACCEPTED MANUSCRIPT 1 Under the controller in [2 4] Under the presented contro ller 0.6 CR IP T Tra cking error ỹ 0.8 0.4 0.2 −0.2 0 AN US 0 5 10 tk 15 20 Figure 4: Tracking error e(tk ). M with the control case when stochastic process and parametric perturbation existing in the tracking control systems. ED Acknowledgment References References PT This work was supported by the Harbin Applied Technology Research and Development Project (excellent academic leader) (No. 2014RFXYJ051), China. AC CE [1] G. Bartolini, A. Ferrara, E. Usai, Output tracking control of uncertain nonlinear second-order systems, Automatica 33 (12) (1997) 2203–2212. [2] T. Fukao, H. Nakagawa, N. Adachi, Adaptive tracking control of a nonholonomic mobile robot, IEEE Transactions on Robotics and Automation 16 (5) (2000) 609–615. [3] H. A. Yousef, M. Hamdy, M. Shafiq, Flatness-based adaptive fuzzy output tracking excitation control for power system generators, Journal of the Franklin Institute 350 (8) (2013) 2334–2353. [4] X. Hu, L. Wu, C. Hu, H. Gao, Adaptive sliding mode tracking control for a flexible air-breathing hypersonic vehicle, Journal of the Franklin Institute 349 (2) (2012) 559–577. [5] Q. Zhou, Y. Gao, H. Li, H. R. Karimi, Control design for discrete-time fuzzy systems with disturbance inputs via delta operator approach, Mathematical Problems in Engineering 724918 (2013) 1–13. [6] T. Takagi, M. Sugeno, Fuzzy identification of systems and its applications to modeling and control, IEEE Transactions on Systems, Man, and Cybernetics 15 (1) (1985) 116–132. [7] K. Tanaka, T. Ikeda, H. O. Wang, Fuzzy regulators and fuzzy observers: relaxed stability conditions and LMI-based designs, IEEE Transactions on Fuzzy Systems 6 (2) (1998) 250–265. [8] J. Qiu, G. Feng, H. Gao, Static-output-feedback control of continuous-time T-S fuzzy affine systems via piecewise lyapunov functions, IEEE Transactions on Fuzzy Systems 21 (2) (2013) 245–261. [9] H. R. Hassanzadeh, M.-R. Akbarzadeh-T, A. Akbarzadeh, A. Rezaei, An interval-valued fuzzy controller for complex dynamical systems with application to a 3-PSP parallel robot, Fuzzy sets and systems 235 (2014) 83–100. 14 ACCEPTED MANUSCRIPT AC CE PT ED M AN US CR IP T [10] Z.-G. Wu, S. Dong, H. Su, C. Li, Asynchronous dissipative control for fuzzy Markov jump systems, IEEE Transactions on Cyberneticsdoi:10.1109/TCYB.2017.2739754. [11] N. Vafamand, M. H. Asemani, A. Khayatian, Robust L1 observer-based non-PDC controller design for persistent bounded disturbed T-S fuzzy systems, IEEE Transactions on Fuzzy Systems. [12] X. Zhang, Y. Han, L. Wu, Y. Wang, State estimation for delayed genetic regulatory networks with reaction-diffusion terms, IEEE Transactions on Neural Networks and Learning Systems 29 (2) (2018) 299–309. [13] X. Zhang, X. Fan, L. Wu, Reduced-and full-order observers for delayed genetic regulatory networks, IEEE Transactions on Cyberneticsdoi:10.1109/TCYB.2017.2726015. [14] R. H. Middleton, G. C. Goodwin, Improved finite word length characteristics in digital control using delta operators, IEEE Transactions on Automatic Control 31 (11) (1986) 1015–1021. [15] R. H. Middleton, G. C. Goodwin, Digital control and estimation: a unified approach, Vol. 56, Prentice Hall Englewood Cliffs, NJ, 1990. [16] L. Guo, M. Tomizuka, High-speed and high-precision motion control with an optimal hybrid feedforward controller, IEEE/ASME Transactions On Mechatronics 2 (2) (1997) 110–122. [17] G. Goodwin, R. L. Leal, D. Mayne, R. Middleton, Rapprochement between continuous and discrete model reference adaptive control, Automatica 22 (2) (1986) 199–207. [18] S. Chen, R. Istepanian, J. Wu, J. Chu, Comparative study on optimizing closed-loop stability bounds of finite-precision controller structures with shift and delta operators, Systems & Control Letters 40 (3) (2000) 153–163. [19] C. P. Neuman, Transformations between delta and forward shift operator transfer function models, IEEE Transactions on Systems, Man and Cybernetics 23 (1) (1993) 295–296. [20] H. Yang, Q. Geng, Y. Xia, L. Li, Output regulation for linear delta operator systems subject to actuator saturation, International Journal of Robust and Nonlinear Control 27 (7) (2017) 1043–1063. [21] Y. Xia, M. Fu, H. Yang, G.-P. Liu, Robust sliding-mode control for uncertain time-delay systems based on delta operator, IEEE Transactions on Industrial Electronics 56 (9) (2009) 3646–3655. [22] H. Li, Y. Gao, L. Wu, H.-K. Lam, Fault detection for TS fuzzy time-delay systems: delta operator and input-output methods, IEEE Transactions on Cybernetics 45 (2) (2015) 229–241. [23] Y. Yuan, P. Zhang, Z. Wang, L. Guo, H. Yang, Active disturbance rejection control for the ranger neutral buoyancy vehicle: A delta operator approach, IEEE Transactions on Industrial Electronics 64 (12) (2017) 9410–9420. [24] Y. Gao, P. Shi, H. Li, S. K. Nguang, Output tracking control for fuzzy delta operator systems with time-varying delays, Journal of the Franklin Institute 352 (7) (2015) 2951–2970. [25] H. Yang, P. Shi, J. Zhang, J. Qiu, Robust H∞ control for a class of discrete time fuzzy systems via delta operator approach, Information Sciences 184 (1) (2012) 230–245. [26] H. Yang, X. Li, Z. Liu, L. Zhao, Robust fuzzy-scheduling control for nonlinear systems subject to actuator saturation via delta operator approach, Information Sciences 272 (2014) 158–172. [27] M. Donkers, W. Heemels, Output-based event-triggered control with guaranteed L∞ -gain and improved and decentralized event-triggering, IEEE Transactions on Automatic Control 57 (6) (2012) 1362–1376. [28] J. I. Poveda, A. R. Teel, A robust event-triggered approach for fast sampled-data extremization and learning, IEEE Transactions on Automatic Control 62 (10) (2017) 4949–4964. [29] C. Peng, Q.-L. Han, D. Yue, To transmit or not to transmit: a discrete event-triggered communication scheme for networked Takagi–Sugeno fuzzy systems, IEEE Transactions on Fuzzy Systems 21 (1) (2013) 164–170. [30] Z.-G. Wu, Y. Xu, R. Lu, Y. Wu, T. Huang, Event-triggered control for consensus of multiagent systems with fixed/switching topologies, IEEE Transactions on Systems, Man, and Cybernetics: Systems. [31] D. Yue, E. Tian, Q.-L. Han, A delay system method for designing event-triggered controllers of networked control systems, IEEE Transactions on Automatic Control 58 (2) (2013) 475–481. [32] Q. Liu, Z. Wang, X. He, D. Zhou, Event-based recursive distributed filtering over wireless sensor networks, IEEE Transactions on Automatic Control 60 (9) (2015) 2470–2475. [33] Y. Fan, L. Liu, G. Feng, Y. Wang, Self-triggered consensus for multi-agent systems with Zeno-free triggers, IEEE Transactions on Automatic Control 60 (10) (2015) 2779–2784. [34] V. Dolk, D. P. Borgers, W. Heemels, Output-based and decentralized dynamic event-triggered control with guaranteed lp -gain performance and Zeno-freeness, IEEE Transactions on Automatic Control 62 (1) (2017) 34–49. [35] C. De Persis, R. Sailer, F. Wirth, Parsimonious event-triggered distributed control: A Zeno free approach, Automatica 49 (7) (2013) 2116–2124. [36] J. Qiu, Y. Xia, H. Yang, J. Zhang, Robust stabilisation for a class of discrete-time systems with time-varying delays via delta operators, IET Control Theory & Applications 2 (1) (2008) 87–93. [37] C. A. Desoer, M. Vidyasagar, Feedback systems: input-output properties, Vol. 55, SIAM, 1975. [38] G. Zames, On the input-output stability of time-varying nonlinear feedback systems part one: Conditions derived using concepts of loop gain, conicity, and positivity, IEEE Transactions on Automatic Control 11 (2) (1966) 228–238. [39] K. Zhou, J. C. Doyle, K. Glover, et al., Robust and optimal control, Vol. 40, Prentice Hall New Jersey, 1996. [40] L. Wu, X. Su, P. Shi, J. Qiu, A new approach to stability analysis and stabilization of discrete-time T-S fuzzy time-varying delay systems, IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics 41 (1) (2011) 273–286. [41] K. Gu, Y. Zhang, S. Xu, Small gain problem in coupled differential-difference equations, time-varying delays, and direct Lyapunov method, International Journal of Robust and Nonlinear Control 21 (4) (2011) 429–451. [42] L. Wu, X. Yang, H.-K. Lam, Dissipativity analysis and synthesis for discrete-time T-S fuzzy stochastic systems with time-varying delay, IEEE Transactions on Fuzzy Systems 22 (2) (2014) 380–394. [43] K.-Y. Lian, J.-J. Liou, Output tracking control for fuzzy systems via output feedback design, IEEE Transactions on Fuzzy 15 ACCEPTED MANUSCRIPT AC CE PT ED M AN US CR IP T Systems 14 (5) (2006) 628–639. 16

1/--страниц