Adaptive Suboptimal Control of Some Discrete-Time Plants with Nonstochastic Bounded Disturbances

The paper addresses adaptive suboptimal control of linear, discrete-time, stationary, minimum-phase, scalar plants in the presence of nonstochastic bounded unmeasurable disturbances with asymmetric upper and lower bounds assumed to be unknown a priori. An additional assumption is that the order of t...

Full description

Saved in:
Bibliographic Details
Published in:Проблеми керування та інформатики
Date:2024
Main Authors: Azarskov, V.M., Solovchuk, K.Y., Volkov, O.Y., Zhiteckii, L.S.
Format: Article
Language:English
Published: Інститут кібернетики ім. В.М. Глушкова НАН України 2024
Subjects:
Online Access:https://nasplib.isofts.kiev.ua/handle/123456789/211146
Tags: Add Tag
No Tags, Be the first to tag this record!
Journal Title:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Cite this:Adaptive Suboptimal Control of Some Discrete-Time Plants with Nonstochastic Bounded Disturbances / V.M. Azarskov, K.Y. Solovchuk, O.Y. Volkov, L.S. Zhiteckii // Проблеми керування та інформатики. — 2024. — № 2. — С. 20–32. — Бібліогр.: 23 назви. — англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1859866275339042816
author Azarskov, V.M.
Solovchuk, K.Y.
Volkov, O.Y.
Zhiteckii, L.S.
author_facet Azarskov, V.M.
Solovchuk, K.Y.
Volkov, O.Y.
Zhiteckii, L.S.
citation_txt Adaptive Suboptimal Control of Some Discrete-Time Plants with Nonstochastic Bounded Disturbances / V.M. Azarskov, K.Y. Solovchuk, O.Y. Volkov, L.S. Zhiteckii // Проблеми керування та інформатики. — 2024. — № 2. — С. 20–32. — Бібліогр.: 23 назви. — англ.
collection DSpace DC
container_title Проблеми керування та інформатики
description The paper addresses adaptive suboptimal control of linear, discrete-time, stationary, minimum-phase, scalar plants in the presence of nonstochastic bounded unmeasurable disturbances with asymmetric upper and lower bounds assumed to be unknown a priori. An additional assumption is that the order of the difference equation describing the plant is known beforehand. The unique feature of the stated problem is that neither the bounds on unmeasured disturbances nor the admissible region to which the unknown parameters of the plant belong are known a priori. Adaptive point and membership set estimation procedures are used to solve this problem Розглядається адаптивне субоптимальне керування лінійними, дискретними, стаціонарними, мінімально-фазовими, скалярними об’єктами за наявності нестохастичних обмежених невимірюваних збурень, верхня та нижня межі яких можуть бути асиметричними та вважаються апріорі невідомими. Додаткове припущення полягає в тому, що порядок різницевого рівняння, яке описує об’єкт, апріорі відомий. Відмінною особливістю поставленої в статті задачі є те, що межі ані невимірюваних збурень, ані допустимої області, до якої належать невідомі параметри об’єкта, не вважаються апріорі відомими. Для вирішення цієї задачі використовуються адаптивні процедури точкового і множинного оцінювання.
first_indexed 2026-03-17T00:18:18Z
format Article
fulltext © V. AZARSKOV, K. SOLOVCHUK, O. VOLKOV, L. ZHITECKII, 2024 20 ISSN 2786-6491 UDC 681.5 V. Azarskov, K. Solovchuk, O. Volkov, L. Zhiteckii ADAPTIVE SUBOPTIMAL CONTROL OF SOME DISCRETE-TIME PLANTS WITH NONSTOCHASTIC BOUNDED DISTURBANCES Valerii Azarskov National Aviation University, Kyiv, orcid: 0009-0004-9206-2475 azarskov@nau.edu.ua Klavdiia Solovchuk International Research and Training Center for Information Technologies and Systems of the National Academy of Sciences of Ukraine and the Ministry of Education and Scie- nce of Ukraine, Kyiv, orcid: 0009-0000-5191-939X solovchuk.ok@gmail.com Oleksandr Volkov International Research and Training Center for Information Technologies and Systems of the National Academy of Sciences of Ukraine and the Ministry of Education and Scie- nce of Ukraine, Kyiv, orcid: 0000-0002-5418-6723 director@irtc.org.ua Leonid Zhiteckii International Research and Training Center for Information Technologies and Systems of the National Academy of Sciences of Ukraine and the Ministry of Education and Scie- nce of Ukraine, Kyiv, orcid: 0000-0002-4560-5113 leonid_zhiteckii@i.ua This paper deals with the adaptive suboptimal control of linear, discrete-time, time-invariant, minimum phase, scalar plants in the presence of nonstochastic bounded unmeasurable disturbances whose upper and lower bounds, which may be asymmetric, are assumed to be unknown a priori. Additional assumption is that an order of the difference equation describing the plant is known a priori. The dis- tinguishing feature of the problem stated in this paper is that neither bounds on the unmeasured disturbances, nor bounds on an allowable region to which the un- known plant parameters belong are assumed to be known a priori. To solve this problem, adaptation procedures for the point and membership set estimation are utilized. The standard recursive procedure with adjustable dead zone is employed in order to derive the point estimates of unknown plant parameters together with the point estimate of time-invariant disturbance component. The size of this dead zone depends on the previous point estimate of the bounds on the time-varying disturbance component and also on a fixed suboptimality index chosen by the de- signer. The estimates generated by the point estimation procedure are directly ex- ploited to derive the adaptive control law. The main idea advanced in this paper is that, instead of unknown a priori membership set of these parameters, their peculi- ar hypothetical a posteriori membership sets are designed via the use of the meas- Міжнародний науково-технічний журнал Проблеми керування та інформатики, 2024, № 2 21 ured system’s signals together with the current point estimate of bounds on the time-varying disturbance component. Contrary to the usual membership set esti- mation approach, this set is updated if only it is discovered that the unknown pa- rameter vector does not belong in reality to this set. To this end, a remarkable property of the point estimation procedure is utilized. Such an approach makes it possible to reconstruct this set and to update the previous estimate of the bounds on the time-varying disturbance component. The finite convergence of the adaptation procedures and also the ultimate boundedness of system’s signals are established. To demonstrate an efficiency of the adaptive controller and support the theoretical study, simulation results are presented. Keywords: adaptive control, point estimation, membership set estimation, dis- crete-time, nonstochastic disturbance. Introduction The problem of a perfect performance of modern control systems with parametric and nonparametric uncertainties remains an actual problem from both theoretical and practical points of view. Within the framework of this scientific problem, novel ideas have been advanced in recent work [1] dealing with the model predictive control ex- tended last time in the modern control theory [2]. An essential progress has been achieved by many researches using adaptive control approaches presented in [3–5]. Im- portant theoretical results have been reported in numerous papers and generalized in several books including, in particular, [6–9]. Some practical application of these adap- tive control approaches for a manufacturing systems with uncertainties has been pre- sented in [10]. There are exist stochastic and nonstochastic approaches to the adaptive control design in the presence of unmeasurable external disturbances. The nonstochastic approach has been advanced by V.M. Kuntsevich who studied methods for the so- called membership set estimation of unknown plant parameters in his book [11] and also by V.A. Yakubovich dealing with the traditional point estimation of these pa- rameters in the book [12, chap. 4]. Their works have been extended for over last two decades in [13, 14]. Recent results achieved in this scientific field can be found in the papers [4, 15]. A distinguishing feature of the disturbances considered in the works above men- tioned is that they have no probabilistic nature [12, p. 137]. Namely, their sequences do not have any stochastic characteristics such as the expected value, variance and others. Nevertheless, it is known [12, item 4.2.1] that, in contrast to stochastic case, the exact identification of a plant subjected to unmeasured nonstochastic disturbances is impossi- ble, in principle, even when the asymptotic behavior of a plant to be controlled is con- sidered. One of the areas which attracts an attention of the adaptive control community up-to-date remains the rejection of arbitrary bounded nonstochastic disturbances. Meanwhile, traditional techniques for rejecting random disturbances based on stochastic approaches become here unacceptable. The adaptive optimal control of the linear discrete-time systems without a pure delay in the presence of arbitrary bounded disturbances with known bounds has originally been proposed in the paper [16]. To establish the asymptotic properties of the adaptive closed- loop control system, the Key Technical Lemma reported in [17, p. 181] is utilized. The adaptive suboptimal controller for rejecting nonstochastic bounded disturbance has been proposed before by V.A. Yakubovich together with his disciple V.A. Bondarko utilizing their Frequency Theorem [12, sec. 4.2]. It turns out that adaptive optimal and suboptimal controls based on the point estimation without the membership set estimation are possible if only the time delay is absent even when bounds on the disturbances are known. 22 ISSN 2786-6491 The method for adaptive robust control of the systems with bounded disturbances whose bounds are unknown has been developed by G. Feng in [18] who has estab- lished the ultimate boundedness of all signals in the closed-loop control system, as- suming that a lower bound on the high frequency gain of the open-loop system and al- so its sign are known a priori. However, no suboptimality of his adaptive controller has been achieved. To cope with the absence of knowledge concerning the bounds on these disturbances, V.A. Bondarko devised the adaptive suboptimal control algorithm using a remarkable property of the finitely convergent estimation procedure examined in [12, chap. 2]; see [19]. In order to implement this algorithm, certain knowledge of a priori bounded membership set of unknown plant parameters was needed. The adaptive suboptimal feedback controller for the first-order nonlinear plant in the pres- ence of bounded disturbances with unknown bounds is designed in [13]. Fruitful ideas of this paper have been extended in [14], where the problem of adaptive suboptimal control of the first-order plant with two unknown parameters and unmodelled dynamics is studied. One of the basic assumptions introduced in [13, 14] is that some constraints on the allowable values of unknown parameters are known a priori. A disadvantage of the works [13, 14, 18, 19] is that the bounds on the disturbance are assumed to be sym- metric. This disadvantage has been overcome in [20]. Unfortunately, some a priori knowledge of the membership set is also required to implement this algorithm. A common feature of the works above mentioned is that they require certain a priori information about a bounded membership set of unknown plant parameters. Difficulties that may arise in practice are how to get suitable estimates of the bounds on admissible parameters which will not be conservative and simultaneous- ly to guarantee that each unknown parameter will lie within corresponding bounds. To the best of our knowledge, the existing works do not provide an answer to this question. This paper proposes a new adaptive suboptimal control method to deal with ar- bitrary bounded disturbances having unknown bounds in the absence of any infor- mation about a priori membership set of unknown parameters. Its original version was presented in [21] at 22nd IFAC World Congress which was held in Yokohama (Japan) in July 2023. 1. Problem statement Let the system to be controlled be a linear discrete-time, time-invariant single-input single-output system with no pure delay described by 1 1 1 1 1( 0),t t n t n t n t n ty a y a y b u b u v b− − − −+ + + = + +  (1) where ,ty tu and tv denote the scalar output, input (control) and unmeasurable dis- turbance, respectively, at time instant t. Using the backward shift operator 1q− defined by 1 1,t tq x x− −= rewrite (1) in the form 1 1( ) ( ) ,t t tA q y B q u v− −= + (2) where 1( )A q− and 1( )B q− are the polynomials of the order n given as 1 1 1( ) 1 ,n nA q a q a q− − −= + + + 1 1 1( ) .n nB q b q b q− − −= + + The following assumptions about the system (2) are made: A1) the integer n is known a priori; A2) the system (1) is the minimum phase plant; A3) the coefficients of 1( )A q− and 1( )B q− are unknown; Міжнародний науково-технічний журнал Проблеми керування та інформатики, 2024, № 2 23 A4) the disturbance tv represents the so-called nonstochastic but bounded varia- ble [11, 12] defined by min maxtv t−         (3) with possibly asymmetric lower and upper bounds, i.e., min max ;−    A5) min and max are unknown a priori. An illustrative example of nonstochastic bounded disturbance tv with asymmetric bounds is depicted in Fig. 1 below. Fig. 1 Remark 1. Note that the open-loop minimum phase system (2) may be unstable, in principle. Besides, 1( )A q− and 1( )B q− may not be relatively prime, i.e., the pole-zero cancellation is valid. See [12, item. 4.2.1]. Denoting by 0y a desired output of the system (1), the regulation problem when 0 consty  will be considered. To this end, introduce the ultimate performance index 0( , ) lim sup | |t t J u v y y → = − depending on the control sequence 1 2: { } , ,tu u u u= = generated by some controller and on the disturbance sequence 1 2: { } , ,tv v v v= = . It is clear that 0( ) inf ( , ) inf lim sup | |t u u t J v J u v e    → = = where  is the standard notation of the space of any bounded sequence, characterizes the optimal asymptotic behavior of the output error 0 t te y y= − (4) for a given disturbance sequence v. Definition [19, 20]. The closed-loop control system is said to be suboptimal with a suboptimality index 0  if the requirement min max 0 : [ , ] lim sup sup ( ) t t v vt e J v   →  +  (5) is satisfied for any positive . Remark 2. In contrast with [12], the expression (5) defines the suboptimality index in the additive but not in multiplicative form, and it is here essential. With assumptions A1) to A5) in mind, the problem is to design the adaptive subop- timal controller guaranteeing the achievement of the control goal (5) for any positive  chosen by the designer. tv max min v t 24 ISSN 2786-6491 2. Optimal nonadaptive control (ideal case) Suppose, for the time being, that the coefficients of 1( )A q− and 1( )B q− are known. To derive the optimal control satisfying the requirement min max 0 : [ , ] lim sup | | sup ( ), t t v vt e J v   →  (6) write the variable tv in the following form: ,t tv v v= + (7) where v and tv are the time-invariant and the time-varying components of ,tv respec- tively. Due to (3) they are given as min max( ) / 2.v =  + , 0,1, 2,tv t  = (8) with max min( ) / 2. =  − (9) Next, substituting (7) into (1) and shifting the indices of ty and ,tu one obtains 1 1 1 1 1 1.t t n t n t n t n ty a y a y b u b u v v+ − + − + ++ + + = + + + (10) Following the standard steps in deriving the optimal control of the discrete-time systems in the presence of bounded disturbances taken, e.g., from [12], one puts 0 1 1 1 1t n t n t n t na y a y b u b u v y− + − +− − − + + + = to get the control law 1 0 1 1 1 2 1 1[ ].t t n t n t n t nu b y a y a y b u b u v− − + − − += + + + − − − (11) By virtue of (10) and the constraints (8), this control law yields 0 ty y t−    implying that the optimality requirement (6) rewritten as 0 : [ , ] lim sup sup ( ), t t v vt e J v  − →  will be achieved with 0 : | |sup ( ) , tv v J v    where (4) has been utilized. 3. Main result Basic ideas. Return to the true situation when the components of the 2n-dimen- sional vector T 1 1[ , , , , , ]n na a b b = and also the (2 1)n + -dimensional extended vector T T[ , ]v =  are unknown. According to the so-called certainty equivalence adaptive control principle [17], replace unknown  by some updated estimate T 1[ ( ), ,t a t = 1( ), ( ), , ( ), ( )]n na t b t b t v t which will be found later. Then from (11), the adaptive control law in the form 1 0 1 1 1 2 1 1( )[ ( ) ( ) ( ) ( ) ( )]t t n t n t n t nu b t y a t y a t y b t u b t u v t− − + − − += + + + − − − − (12) can directly be written. Міжнародний науково-технічний журнал Проблеми керування та інформатики, 2024, № 2 25 Further, recalling that the upper bound,  on { }tv defined in (9) remains unknown (since min and max are unknown), a current estimate, t of  will be exploited to devise the adaptive algorithm for updating both the parameter estimate vector 2 1R n t +  and scalar .t Define some strip tS in the (2 1)thn+ Euclidean space of extended vectors T 1 1[ , , , , , , ]n na a b b v = as T 2 1 1: { : | | } R ,n t t ttS y + −=  −     (13) where TT 1 1 1 1,[ , 1] [ , , , , ,t t t t n ty y u− − − − − =  = − − ,1]t nu − denotes the extended re- gression vector. It follows from (10) together with (8) that tS will contain the unknown vector T T[ , ]v =  if [0, )sup | | .t t tv    It turns out that two different cases shown in Fig. 2 are possible. In the case depict- ed in Fig. 2 left, a current estimate t of unknown upper bound  on the absolute value of the disturbance v is not less than its true value  . Then the unknown extended pa- rameter vector  will belong to the strip defined by the expression (13). However, in the other case depicted in this figure right when ,t   the unknown expended parame- ter vector  may not belong to this strip. Fig. 2 Now, consider some intersection written in the recursive form 1t t tS− =   (14) as depicted in Fig. 3, a, b. a b Fig. 3 tS tS 1t− 1t− t 0 0 tS { } { }   tS 26 ISSN 2786-6491 It can be understood that if 1t− and also tS  then t   (see Fig. 3, a) whereas if 1t− but tS  then t becomes empty (see Fig. 3, b) meaning that the set of the inequalities T 1 0 0 ˆ ( , 1, , )k k ky k k k t−−    = + with respect to unknown  will be incompatible. With this fact in mind, put 1 / 2t t− =  + (15) if it is discovered that ,t =  and 1t t− =  (16) otherwise. It is clear that the increase of the estimates st of the upper bound on | |tv cannot be infinite, since t becomes nonempty if [0, )sup | | .t t tv   To avoid the case ,t =  fix a 0,kt  put 1 2k k kt t t n+ + =  = =  ( 0,1, )k = and consider the intersection [ ]k of the 2 1n + strips tS defined in (13) at the time instants , 1, , 2k k kt t t t n= + + as 1 2[ ] . k k kt t t nk S S S+ + =    (17) Obviously, the set [ ]k defined in (17) represents a convex bounded polytope iff the vectors 2, , k kt t n+  are linearly independent as illustrated in Fig. 4. Its Cheby- shev center c  and radius rad  are specified as follows: c 2 arg min max ,          =  −  c ( ) 2 1, , rad max ,i i N h =  =  − where ( )ih is the notation of ith vertex of  and N is the number of these vertices, and rad  denotes the radius of , 2  is the Euclidean vector norm. Fig. 4 c 2kt nS + kt S [ ]k (1)h (2)h ( )h ( )Nh 2kt n+ kt  Міжнародний науково-технічний журнал Проблеми керування та інформатики, 2024, № 2 27 One of the basic ideas advanced in this paper is to utilize the set [ ]k given by (17) as some hypothetical a posteriori membership set of unknown  since [ ]k may not contain  in reality if kt  is sufficiently small. On the other hand, the condi- tion [ ]k  will be guaranteed via a sequential increase of kt  from a time kt to some time 1kt + for 0,1,k = . (Such an approach will be described in the subsection below.) Adaptive law. The adaptation algorithm is implemented via the following subse- quent steps. Step 1. Starting from 0,k = fix some initial 0 ,t t= put 0 / 2,t =  and find the vertices (1) ( ), , Nh h of [0] using the measured vectors t at 0 0, , 2 .t t t n= + These vectors must be linearly independent to ensure the boundedness of [ ].k To do it, choose arbitrarily the initial control sequence 0 0 01 2, , ,t t t nu u u+ + so that the Gram’s determinants t satisfy 0t  for 0 0[ , 2 ],t t t n + (18) where 0 00 0 0 0 00 0 T T T T det , 1, , 2 . t t jt t t j t j t jt t j j n + + + ++          = =          (Note that the requirement (18) defines the necessary and sufficient condition for the linear independence of these vectors [22, chap. 9, item 3].) Step 2. Compute the Chebyshev center c 2kt n+ using, e.g., [23] and the radius of [ ].k Step 3. Choose the initial estimate 2kt n+ moving it to the Chebyshev center c 2 , kt n+ i.e., set c 2 2 . k kt n t n+ + =  Step 4. Set 0t = at 2 ,kt t n= + where t denotes an auxiliary variable deter- mined below. Step 5. Update t s using the adaptive estimation procedure with the adjustable dead zone as 0 1 1 1 12 1 2 ( , , ) ,t t t t t t t t f e − − − − −    =  −    (19) at 2 1,kt t n + + where 0 / 2,t t =  +  (20) 0 0 0 0 0 if , ( , , ) 0 if (0 ), if e e f e e e e  −        =        +   − denotes the dead-zone function caused by the identification algorithm of [12, chap. 2] and t is the coefficient chosen free from 28 ISSN 2786-6491 0 2t        (21) to ensure 1( ) 0.b t  Step 6. If kt t then calculate t as follows: 0 1 1 1 2 2 2 1 1 21 if ( , , ) 0, (2 )(| | ) || || otherwise. t t t t t t t t t t t f e e − − − − − − −    =   =   +  −  −  (22) Step 7. Verify the condition 2(rad [ ]) .t k   (23) If (23) is violated then put 0t = and increase k by 1 and instead of 1t− the updated t specified in (15) exploit, reconstruct [ ]k using the same initial vectors t given in Step 1 with updated t and return to Step 2. Remark 3. It can be proved that the violation of (23) indicates that [ ]k for given . kt  This remarkable fact makes it possible to discover indirectly that the inter- section (17) is indeed empty without performing complex computations. Convergence and ultimate boundedness analysis. A remarkable property of the procedure (19)–(21) leading to (23) is given in the following lemma: Lemma. If [0, )sup | |t t tv    and c 0 [ 1]k =  − then 1 22 2 0 2 1 1 1 2 (2 ) ( , , ) (rad [ 1]) . k t t t t t tt t f e k − − − − −  −       − Proof. Follows from the features of adaptive estimation procedure established in [12, chap. 2]. □ With this lemma, the ultimate properties of the adaptive controller proposed here is summarized in next theorem. Theorem. Let assumptions A1) to A5) be satisfied. Then the resulting closed-loop control system including the plant (1), the control law (12) and the adaptation algo- rithm described in Steps 1 to 7 has the following properties: i) the adaptive estimation algorithm determined by (15), (16), (19)–(23) together with (13) and (17) converges at a finite time; ii) the system signals are ultimately bounded; iii) the suboptimal control objective (5) is achieved. Proof. Due to space limitation, details are omitted. □ 4. A numerical example and simulations To demonstrate the features of the proposed adaptive controller, the simplest first- order system described by 1 12,4t t t ty y u v− −− = + was considered. (Although the instability is not inherent for many industrial plants to be controlled, firstly, for open process control systems, 1( )A q− was purposely chosen to be unstable (as in [14]), in order to show that even in such a «bad» case, the adaptive suboptimal performance can be achieved.) In this example, the suboptimality index was chosen as 0,1. = Міжнародний науково-технічний журнал Проблеми керування та інформатики, 2024, № 2 29 Before going to simulation experiments, the parameters of sets [ ]k of vectors T 1 1 ˆˆ ˆ[ , ],a b = 0,1, 2, 3k = for t = 0,05; 0,1; 0,15; 0,2 were calculated and summa- rized in the Table. These sets are depicted in Fig. 5, showing that [3] which contains unknown  is an appropriate a posteriori membership set of this vector while [0], [1]  and [2] are not. Table Ordinal number k Chebyshev center Radius of [ ]k 0 [– 2,28, 1,26] 0,092 1 the same 0,126 2 the same 0,181 3 the same 0,320 Results of the simulation experiment conducted for [ 0,2, 0,2]tv  − and 0 5y = are given in Fig. 6. They show the behavior of the adaptive control system in this exam- ple. It can be seen from Fig. 6, a that if the variable t exceeds the square of the radius of [ ]k designed at the time instant kt t= ( 0,1, 2, 3),k = then by virtue of the algo- rithm determined by (15), (16), (19)–(23) together with (13) and (17) the variable t is reset to zero, whereas the current estimate of the bound on disturbance increases with- out exceeding a certain threshold / 2 0, 25 +  = (Fig. 6, b). One observes that the components of T 1 1[ ( ), ( )]t a t b t = jump moving toward the Chebyshev center c at the same time instants kt t= as shown in Fig. 6, c and d. Fig. 5 For comparison, the behavior of nonadaptive control system with the same plant and fixed parameters of the controller specified by c (instead of unknown ) is pre- sented in Fig. 7. It turned out that this control system was also stable and its output error remained bounded. However, one can observe that the performance of the adaptive con- trol system is essentially better than in the nonadaptive case. Moreover, it is seen that the suboptimality property of the proposed adaptive controller is ensured (Fig. 6, e). 1a 1,0 1,2 1,4 1,6 1b̂ 1b c Chebyshev center True parameter vector      – 2,8 – 2,4 – 2,0 1̂a 30 ISSN 2786-6491 Fig. 6 Fig. 7 – 1,2 0 500 1000 1500 e(t) – 0,8 t – 0,6 – 1,0 1500 e 0,00 0 500 1000 1500 t a  t Zeroing  t at t = t0, t1, t2 Zeroing  t at t = t3 0,01 0,02 0,10 0 500 1000 1500 t b 0,03  t Change in estimating  at t = t0, t1, t2 Change in estimating  at t = t3 0,14 – 2,38 0 500 1000 1500 c a 1 (t) Initialization of adaptation process at t = t0, t1, t2 Initialization of adaptation process at t = t3 First component of Chebyshev center at t = t0,…, t3 – 2,34 – 2,30 0,98 0 500 1000 1500 b 1 (t) Initialization of adaptation process at t = t0, t1, t2 Initialization of adaptation process at t = t3 Second component of Chebyshev center at t = t0,…, t3 1,06 d t t – 1,5 0 500 1000 e(t) – 0,5 1,26 t 0,5 Міжнародний науково-технічний журнал Проблеми керування та інформатики, 2024, № 2 31 Conclusion The adaptive suboptimal discrete-time closed-loop system containing the linear uncertain scalar minimum phase plant subjected to nonstochastic bounded disturbances with unknown bounds in the absence of any a priori information about the membership set of its parameters is addressed in this paper. The asymptotical features of the adaptive controller above proposed are established. They shed some light on the possibility of achieving any suboptimality index utilizing the adaptive controller. In future, this cont- roller can be improved via the use of additional ellipsoid estimation procedure in order to accelerate the adaptation process. The MIMO case will also be studied. В.М. Азарсков, К.Ю. Соловчук, О.Є. Волков, Л.С. Житецький АДАПТИВНЕ СУБООПТИМАЛЬНЕ КЕРУВАННЯ ДЕЯКИМИ ДИСКРЕТНИМИ ОБ’ЄКТАМИ З НЕСТОХАСТИЧНИМИ ОБМЕЖЕНИМИ ЗБУРЕННЯМИ Азарсков Валерій Миколайович Національний авіаційний університет, м. Київ, azarskov@nau.edu.ua Соловчук Клавдія Юріївна Міжнародний науково-навчальний центр інформаційних технологій та систем НАН України та МОН України, м. Київ, solovchuk.ok@gmail.com Волков Олександр Євгенович Міжнародний науково-навчальний центр інформаційних технологій та систем НАН України та МОН України, м. Київ, director@irtc.org.ua Житецький Леонід Сергійович Міжнародний науково-навчальний центр інформаційних технологій та систем НАН України та МОН України, м. Київ, leonid_zhiteckii@i.ua Розглядається адаптивне субоптимальне керування лінійними, дискретними, стаціонарними, мінімально-фазовими, скалярними об’єктами за наявності не- стохастичних обмежених невимірюваних збурень, верхня та нижня межі яких можуть бути асиметричними та вважаються апріорі невідомими. Додаткове припущення полягає в тому, що порядок різницевого рівняння, яке описує об’єкт, апріорі відомий. Відмінною особливістю поставленої в статті задачі є те, що межі ані невимірюваних збурень, ані допустимої області, до якої належать невідомі параметри об’єкта, не вважаються апріорі відомими. Для вирішення цієї задачі використовуються адаптивні процедури точкового і множинного оцінювання. Стандартна рекурентна процедура з регульованою зоною нечутли- вості застосовується для отримання точкових оцінок невідомих параметрів об’єкта разом з точковою оцінкою постійної компоненти збурення. Розмір цієї зони нечутливості залежить від попередньої точкової оцінки меж змінної у часі компоненти збурення, а також від фіксованого показника субоптимальності, вибраного конструктором системи. Оцінки, отримані за допомогою процедури точкового оцінювання, безпосередньо використовуються для отримання закону адаптивного керування. Головна ідея, висунута в статті, полягає в тому, що за- мість невідомої апріорної множини належності цих параметрів використову- ються їхні своєрідні гіпотетичні апостеріорні множини належності, розроблені mailto:leonid_zhiteckii@i.ua 32 ISSN 2786-6491 за допомогою вимірюваних сигналів системи разом з поточною точковою оцін- кою меж змінної у часі компоненти збурення. На відміну від звичайного підхо- ду до оцінки множини належності цей набір оновлюється лише тоді, коли вияв- ляється, що невідомий вектор параметрів насправді не належить цій множині. Для цього застосовується певна властивість процедури точкового оцінювання. Такий підхід дає змогу реконструювати цю множину і оновити попередню оці- нку меж компоненти збурення, змінної в часі. Встановлено скінченну збіжність процедур адаптації, а також граничну обмеженість сигналів системи. Для де- монстрації ефективності адаптивного регулятора та підтримки теоретичного до- слідження репрезентовані результати моделювання. Ключові слова: адаптивне керування, точкове оцінювання, множинне оці- нювання, дискретний час, нестохастичне збурення. REFERENCES 1. Gubarev V. New trends in control theory. Міжнародний науково-технічний журнал Проблеми керування та інформатики. 2022. N 3. P. 5–21. DOI: https://doi.org/10.34229/2786-6505-2022-3-1 2. Rawlings J.B., Mayne D.G. Model predictive control. Theory and design. Madison : Nob Hill Publishing, 2009. 576 p. 3. Shiota T., Ohmori H. Design of adaptive I-PD control system with variable reference model. Aus- tralian Control Conference. Australia : Fremantle, WA, 2013. P. 115–120. 4. Sokolov V.F. Adaptive stabilization of minimum-phase plant under Lipschitz uncertainty via Ya- kubovich's method of recurrent objective inequalities. IFAC-PapersOnLine. 2015. N 48. P. 563–567. 5. Zhiteckii L.S., Solovchuk K.Yu. Adaptive robust multivariable control of noninvertible memor- yless systems with bounded disturbances: a generalization. Міжнародний науково-технічний журнал Проблеми керування та інформатики. 2022. N 2. P. 22–38. DOI: https://doi.org/10. 34229/2786-6505-2022-2-2 6. Aström K.J., Wittenmark B. Adaptive control: 2nd ed. New York : Dover Publications, 2014. 573 p. 7. Ioannou P., Sun J. Robust adaptive control. Mineola, New York : Dover Publications, 2013. 821 p. 8. Narendra K.S., Annaswamy A.M. Stable adaptive systems. Mineola, New York : Dover Publica- tions, 2012. 512 p. 9. Tao G. Adaptive control design and analysis. New York : John Wiley and Sons, 2013. 640 p. 10. Azarskov V.N., Zhiteckii L.S., Solovchuk K.Yu., Sushchenko O.A., Lupoi R.O. Inventory control for a manufacturing system under uncertainty: adaptive approach. Proc. 20th IFAC World Con- gress. France : Toulouse, 2017. P. 10154–10159. 11. Kuntsevich V.M. Control under uncertainty: guaranteed results in control and identification prob- lems. Kyiv : Nauk. dumka, 2006. 264 p. (in Russian). 12. Fomin V.N., Fradkov A.L., Yakubovich V.A. Adaptive control of dynamic plants. Moscow : Nauka, 1981. 448 p. (in Russian). 13. Sokolov V.F. Adaptive suboptimal tracking for the first-order plant with Lipschitz uncertainty. IEEE Transactions on Automatic Control. 2003. N 48. P. 607–612. 14. Sokolov V.F. Adaptive suboptimal robust control of the first-order plant. Automation and Remote Control. 2008. N 69. P. 1357–1372. 15. Zhiteckii L.S., Solovchuk K.Yu. Robust and robustly-adaptive control of some noninvertible memoryless systems. Recent Developments in Automatic Control Systems (Y.P. Kondratenko, V.M. Kuntsevich, A.A. Chikrii, and V.F. Gubarev, eds), chap. 3. Gistrup, Denmark : River Pub- lishers, 2022. P. 61–96. 16. Ortega R., Lozano-Leal R. A note on direct adaptive control of systems with bounded disturb- ances. Automatica. 1987. N 23. P. 253–254. 17. Goodwin G.C., Sin K.S. Adaptive filtering, prediction and control. Engewood Cliffs, New Jersey : Prentice-Hall, 1984. 540 p. 18. Feng G. A robust discrete-time direct adaptive control algorithm. Systems & Control Letters. 1994. N 22 (3). P. 203–208. 19. Bondarko V.A. Adaptive suboptimal control of solutions of linear discrete equations. Dokl. Akad. Nauk SSSR. 1983. N 270 (2). P. 301–303. (in Russian). 20. Zhiteckij L.S. Adaptive control of systems subjected to bounded disturbances. Bounding Ap- proaches to System Identification (M. Milanese etc., eds.), chap. 24. New York : Plenum Press, 1996. P. 383–407. 21. Azarskov V.N., Solovchuk K.Yu., Volkov O.Ye., Zhiteckii L.S. A new adaptive suboptimal con- troller for a class of linear discrete-time systems with unknown bounded disturbances. Proc. 22nd IFAC World Congress. Yokohama : Japan. 2023. P. 9924–9929. 22. Gantmacher F.R. The theory of matrices. New York : AMS Chelsea Publishing, 1960. 660 p. 23. Botkin N.D., Turova-Botkina V.L. An algorithm for finding the Chebyshev center of a convex polyhedron. Technical Report, Institut für Angewandte Mathematik und Statistik. Würzburg : Jul- ius-Maximilians-Universität, 1992. P. 395. Submitted 02.04.2024 https://jais.net.ua/index.php/files/issue/view/4 https://jais.net.ua/index.php/files/issue/view/4 https://doi.org/10.34229/2786-6505-2022-3-1 https://jais.net.ua/index.php/files/issue/view/4 https://jais.net.ua/index.php/files/issue/view/4 https://doi.org/10.34229/2786-6505-2022-2-2 https://doi.org/10.34229/2786-6505-2022-2-2
id nasplib_isofts_kiev_ua-123456789-211146
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 0572-2691
language English
last_indexed 2026-03-17T00:18:18Z
publishDate 2024
publisher Інститут кібернетики ім. В.М. Глушкова НАН України
record_format dspace
spelling Azarskov, V.M.
Solovchuk, K.Y.
Volkov, O.Y.
Zhiteckii, L.S.
2025-12-24T21:17:43Z
2024
Adaptive Suboptimal Control of Some Discrete-Time Plants with Nonstochastic Bounded Disturbances / V.M. Azarskov, K.Y. Solovchuk, O.Y. Volkov, L.S. Zhiteckii // Проблеми керування та інформатики. — 2024. — № 2. — С. 20–32. — Бібліогр.: 23 назви. — англ.
0572-2691
https://nasplib.isofts.kiev.ua/handle/123456789/211146
681.5
10.34229/1028-0979-2024-2-2
The paper addresses adaptive suboptimal control of linear, discrete-time, stationary, minimum-phase, scalar plants in the presence of nonstochastic bounded unmeasurable disturbances with asymmetric upper and lower bounds assumed to be unknown a priori. An additional assumption is that the order of the difference equation describing the plant is known beforehand. The unique feature of the stated problem is that neither the bounds on unmeasured disturbances nor the admissible region to which the unknown parameters of the plant belong are known a priori. Adaptive point and membership set estimation procedures are used to solve this problem
Розглядається адаптивне субоптимальне керування лінійними, дискретними, стаціонарними, мінімально-фазовими, скалярними об’єктами за наявності нестохастичних обмежених невимірюваних збурень, верхня та нижня межі яких можуть бути асиметричними та вважаються апріорі невідомими. Додаткове припущення полягає в тому, що порядок різницевого рівняння, яке описує об’єкт, апріорі відомий. Відмінною особливістю поставленої в статті задачі є те, що межі ані невимірюваних збурень, ані допустимої області, до якої належать невідомі параметри об’єкта, не вважаються апріорі відомими. Для вирішення цієї задачі використовуються адаптивні процедури точкового і множинного оцінювання.
en
Інститут кібернетики ім. В.М. Глушкова НАН України
Проблеми керування та інформатики
Адаптивне керування та методи ідентифікації
Adaptive Suboptimal Control of Some Discrete-Time Plants with Nonstochastic Bounded Disturbances
Адаптивне субоптимальне керування деякими дискретними об’єктами з нестохастичними обмеженими збуреннями
Article
published earlier
spellingShingle Adaptive Suboptimal Control of Some Discrete-Time Plants with Nonstochastic Bounded Disturbances
Azarskov, V.M.
Solovchuk, K.Y.
Volkov, O.Y.
Zhiteckii, L.S.
Адаптивне керування та методи ідентифікації
title Adaptive Suboptimal Control of Some Discrete-Time Plants with Nonstochastic Bounded Disturbances
title_alt Адаптивне субоптимальне керування деякими дискретними об’єктами з нестохастичними обмеженими збуреннями
title_full Adaptive Suboptimal Control of Some Discrete-Time Plants with Nonstochastic Bounded Disturbances
title_fullStr Adaptive Suboptimal Control of Some Discrete-Time Plants with Nonstochastic Bounded Disturbances
title_full_unstemmed Adaptive Suboptimal Control of Some Discrete-Time Plants with Nonstochastic Bounded Disturbances
title_short Adaptive Suboptimal Control of Some Discrete-Time Plants with Nonstochastic Bounded Disturbances
title_sort adaptive suboptimal control of some discrete-time plants with nonstochastic bounded disturbances
topic Адаптивне керування та методи ідентифікації
topic_facet Адаптивне керування та методи ідентифікації
url https://nasplib.isofts.kiev.ua/handle/123456789/211146
work_keys_str_mv AT azarskovvm adaptivesuboptimalcontrolofsomediscretetimeplantswithnonstochasticboundeddisturbances
AT solovchukky adaptivesuboptimalcontrolofsomediscretetimeplantswithnonstochasticboundeddisturbances
AT volkovoy adaptivesuboptimalcontrolofsomediscretetimeplantswithnonstochasticboundeddisturbances
AT zhiteckiils adaptivesuboptimalcontrolofsomediscretetimeplantswithnonstochasticboundeddisturbances
AT azarskovvm adaptivnesuboptimalʹnekeruvannâdeâkimidiskretnimiobêktamiznestohastičnimiobmeženimizburennâmi
AT solovchukky adaptivnesuboptimalʹnekeruvannâdeâkimidiskretnimiobêktamiznestohastičnimiobmeženimizburennâmi
AT volkovoy adaptivnesuboptimalʹnekeruvannâdeâkimidiskretnimiobêktamiznestohastičnimiobmeženimizburennâmi
AT zhiteckiils adaptivnesuboptimalʹnekeruvannâdeâkimidiskretnimiobêktamiznestohastičnimiobmeženimizburennâmi