Professional Documents
Culture Documents
2013 - Chen&Yamada - A New Design Method of High-Order Modified Repetitive Control Systems For Reference Inputs With Uncertain Period-Time PDF
2013 - Chen&Yamada - A New Design Method of High-Order Modified Repetitive Control Systems For Reference Inputs With Uncertain Period-Time PDF
Research Article
A New Design Method of High-Order Modified
Repetitive Control Systems for Reference Inputs with
Uncertain Period-Time
Copyright © 2013 Zhongxiang Chen et al. This is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
cited.
This paper considers the design of high-order modified repetitive control systems for periodic reference inputs with uncertain
period-time. The objective of this work was to develop a new design method so that the closed-loop high-order modified
repetitive control system is robustly stable with high control precision for periodic reference inputs with uncertain period-
time. The parametrization of all stabilizing controllers containing three free parameters is proposed based on the Youla-Kucera
parameterization. Moreover, to obtain the free parameters, the constraint conditions were converted into stability conditions in
the form of Bilinear Matrix Inequalities that can be solved using the available software. In addition, the high control precision is
guaranteed by designing the free parameters after the control characteristic of this control system. The validity and effectiveness of
the proposed design method were verified by numerical examples.
For multiple-period systems that have a similar structure to In Figure 1, the structure of the high-order modified repet-
the high-order repetitive control system, solutions are given itive controller 𝐶𝑟 (𝑠) consists of multiple periodic signal
by Yamada et al. [12] and Yamada et al. [13] using the Youla- generators and can be expressed as
Kucera parametrization (YKP) without solving the problem
of designing the free parameters. 𝐶𝑟 (𝑠) = 𝐻 (𝑠) (𝐼 − 𝐻 (𝑠))−1 = (𝐼 − 𝐻 (𝑠))−1 𝐻 (𝑠) , (2)
In this paper, we consider the problem of designing a 𝑁
high-order modified repetitive control system to track or 𝐻 (𝑠) = ∑𝑊𝑖 𝑞𝑖 (𝑠) 𝑒−𝑠𝐿 𝑖 , (3)
reject external signals with uncertain period-time. 𝑖=1
A new structure of high-order modified repetitive con-
troller was established inspired by the conventional high- 𝐿 𝑖 = 𝑖 × 𝐿, (4)
order repetitive controller. Two useful methodologies, the where 𝑊𝑖 (𝑖 = 1, . . . , 𝑁) are the weighting factors to modify
YKP and the Bilinear Matrix Inequalities (BMIs) approach, the dynamic response in between the harmonic frequencies,
were adopted to design compensators in the high-order mod- and 𝑞(𝑠) ∈ 𝑅𝑝×𝑝 (𝑠) is a low-pass filter assumed as a diagonal
ified repetitive control system. The objective was to develop a matrix function.
new and simple design method that not only provides good Like the current high-order repetitive control design
control performance while preserving the system’s stability, approaches in [5, 7, 10], this work relied on the following two
but can also be efficiently implemented with available soft- nonrestrictive assumptions.
ware. The approach taken in this paper was as follows. First,
the parametrization of all stabilizing controllers was derived Assumption 1. All harmonics 𝑙 = 0, . . . , ℎ of the reference
on the basis of the YKP. Next, the constraint conditions of the input are assumed to lie in the passband of 𝑞(𝑠). This can
free parameters were transformed into stability conditions be guaranteed by the proper design of 𝑞(𝑠), provided that all
in the form of BMIs. In addition, according to the control harmonics 𝑙 = 0, . . . , ℎ lie well below 𝜔𝑐𝑜 , the cutoff frequency
characteristic, a low-pass filter was introduced to specify of 𝑞(𝑠).
the intermediate frequencies. Finally, numerical examples
for a single-input/single-output (SISO) tracking system were Assumption 2. In its passband, 𝑞(𝑠) is assumed to equal its DC
specified to illustrate the validity and effectiveness of the gain 𝑞(𝑗𝜔) = 𝐼 for all 𝜔 < 𝜔𝑐𝑜 . These two assumptions imply
proposed design method. The results presented in this paper that
show that the high-order modified repetitive control system
𝐶𝑟 (𝑠) = 𝐶 ̂ (𝑠))−1 𝐻
̂𝑟 (𝑠) := (𝐼 − 𝐻 ̂ (𝑠) , (5)
can provide high control precision for periodic signals with
a certain vibration in period-time. Moreover, the control 𝑁
precision at the intermediate frequencies was satisfied by ̂ (𝑠) = ∑𝑊𝑖 𝑒−𝑠𝐿 𝑖 ,
𝐻 (6)
modifying the cutoff frequency of low-pass filters in compen- 𝑖=1
sators.
for all 𝑠 = 𝑗𝜔𝑙 (𝑙 = 0, . . . , ℎ), where 𝐿 𝑖 (𝑖 = 1, . . . , 𝑁) are
Notations. 𝑅𝐻∞ denotes the set of stable proper real rational defined in (4) and 𝜔𝑙 (𝑙 = 0, . . . , ℎ) are the frequency com-
functions; 𝐻∞ denotes the set of stable causal functions; 𝑅𝑛 ponents of the periodic reference input described as
denotes the 𝑛-dimensional Euclidean space; 𝑅𝑛 × 𝑚 is the set 2𝑙𝜋
of all 𝑛 × 𝑚 real matrices; 𝐼 is the identity matrix; the null 𝜔𝑙 := , 𝑙 = 0, . . . , ℎ. (7)
𝐿
matrix or null vector of appropriate dimension is denoted by
0; the superscript “𝑇” stands for the transpose of a matrix; In this case, the 𝑁 weighting factors 𝑊𝑖 (𝑖 = 1, . . . , 𝑁) can be
𝜎{⋅} denotes the largest singular value of {⋅}; the symmetric determined by using the methods proposed in [5, 7, 10].
terms in a symmetric matrix are denoted by ∗; for example, In addition, suppose that the strictly proper plant, 𝐺(𝑠), is
controllable and observable and its state-space description is
[ 𝐴∗ 𝐶𝐵 ] = [ 𝐵𝐴𝑇 𝐶𝐵 ].
𝑥̇ (𝑡) = 𝐴 𝑝 𝑥 (𝑡) + 𝐵𝑝 𝑢 (𝑡) ,
(8)
2. Problem Formulation 𝑦 (𝑡) = 𝐶𝑝 𝑥 (𝑡) ,
Figure 1 depicts the proposed high-order modified repetitive where 𝐴 𝑝 ∈ 𝑅𝑛×𝑛 , 𝐵𝑝 ∈ 𝑅𝑛×𝑚 , and 𝐶𝑝 ∈ 𝑅𝑝×𝑛 are system
control system, where 𝐺(𝑠) ∈ 𝑅𝑝 ×𝑚 (𝑠) is a strictly proper matrices, and 𝑥(𝑡) ∈ 𝑅𝑛 , 𝑢(𝑡) ∈ 𝑅𝑚 , and 𝑦(𝑡) ∈ 𝑅𝑝 denote
plant with an assumption that 𝑝 ≤ 𝑚. 𝐶1 (𝑠) ∈ 𝑅𝑚×𝑝 (𝑠) state, control input, and control output, respectively.
is a forward compensator to enhance the robust stability of The design problem considered in this paper can be stated
this closed-loop system and improve the transient response, as follows: for a given strictly proper plant 𝐺(𝑠), a constant 𝑁,
𝐶2 (𝑠) ∈ 𝑅𝑚×𝑝 (𝑠) is typically called the learning filter and and a low-pass filter 𝑞(𝑠), find the admissible compensators
is used to compensate the high-order modified repetitive 𝐶1 (𝑠) and 𝐶2 (𝑠) such that the controller 𝐶(𝑠) written as
controller 𝐶𝑟 (𝑠), 𝑦(𝑡) ∈ 𝑅𝑝 is the control output, and 𝑟(𝑡) ∈
𝑅𝑝 is a periodic reference input. Uncertainty of period-time 𝐶 (𝑠) = 𝐶1 (𝑠) + 𝐶2 (𝑠) 𝐶𝑟 (𝑠) (9)
is modeled as relative uncertainty on 𝐿, bounded by Δ:
robustly stabilizes the closed-loop system with high control
precision for the periodic reference input with uncertainty in
𝐿 𝛿 := 𝐿 (1 + 𝛿) , |𝛿| ≤ Δ < ∞. (1) period-time, where 𝐶𝑟 (𝑠) is defined in (2).
Mathematical Problems in Engineering 3
3. High-Order Modified Repetitive Control 𝐶(𝑠) can be written as (10a) and (10b). Substitute 𝐶(𝑠) in (9)
System Design into (10a) and the 𝑄(𝑠) in (11) can be represented as
3.1. The Parametrization of All Stabilizing Controllers. In this 𝑄 (𝑠) = {(𝑌 (𝑠) 𝐶1 (𝑠) − 𝑋 (𝑠))
section, we derive the parametrization of all stabilizing con-
trollers, with the structure of the high-order modified repet- + (𝑌 (𝑠) 𝐶2 (𝑠) − 𝑌 (𝑠) 𝐶1 (𝑠) + 𝑋 (𝑠)) 𝐻 (𝑠)}
itive controller for periodic reference inputs with uncertain
̃ (𝑠) + 𝑁
× {(𝐷 ̃ (𝑠) 𝐶1 (𝑠))
period-time. The necessary and sufficient conditions for the
parametrization of all stabilizing controllers by means of the −1
YKP are proposed by the following main theorem. + (𝑁 ̃ (𝑠) 𝐶1 (𝑠) − 𝐷
̃ (𝑠) 𝐶2 (𝑠) − 𝑁 ̃ (𝑠)) 𝐻 (𝑠)} .
(13)
Theorem 1. The control system in Figure 1 is internally stable
if and only if the controllers can be written as Let 𝐶1𝑛 (𝑠) ∈ 𝑅𝐻∞ and 𝐶1𝑑 (𝑠) ∈ 𝑅𝐻∞ be coprime factors of
𝐶1 (𝑠) on 𝑅𝐻∞ satisfying
−1
̃ (𝑠)) (𝑋 (𝑠) + 𝑄 (𝑠) 𝐷
𝐶 (𝑠) = {(𝑌 (𝑠) − 𝑄 (𝑠) 𝑁 ̃ (𝑠)) :
−1
𝐶1 (𝑠) = 𝐶1𝑛 (𝑠) 𝐶1𝑑 (𝑠) . (14)
̃ (∞) ≠ 0}
𝑌 (∞) − 𝑄 (∞) 𝑁 Under the assumptions in Section 2, the low-pass filter 𝑞(𝑠)
(10a) is a diagonal matrix function; this means that 𝐻(𝑠) is also a
diagonal matrix function, and substituting (14) into (13) yields
−1
̃ (𝑠) + 𝐷 (𝑠) 𝑄 (𝑠)) (𝑌
= {(𝑋 ̃ (𝑠) − 𝑁 (𝑠) 𝑄 (𝑠)) :
𝑄 (𝑠) = {(𝑌 (𝑠) 𝐶1𝑛 (𝑠)−𝑋 (𝑠) 𝐶1𝑑 (𝑠))
̃
𝑌 (∞) − 𝑁 (∞) 𝑄 (∞) ≠ 0} , +(𝑌 (𝑠) 𝐶2 (𝑠) 𝐶1𝑑 (𝑠) − 𝑌 (𝑠) 𝐶1𝑛 (𝑠)
(10b)
+𝑋 (𝑠) 𝐶1𝑑 (𝑠)) 𝐻 (𝑠)}
̃ (𝑠) 𝐶1𝑑 (𝑠)+ 𝑁
̃ (𝑠) 𝐶1𝑛 (𝑠)) (15)
̃ (𝑠) − 𝑄0𝑛 (𝑠)) 𝐻 (𝑠)}
𝑄 (𝑠) = {𝑄0𝑛 (𝑠) + (𝑌 (𝑠) 𝑄 ×{(𝐷
matrix function. According to (2), (9), and (13), 𝑄(𝑠) can be description. According to (2) and (11), 𝑄(𝑠) can be expressed
rewritten as as
̃−1 (𝑠) . (18)
𝑄 (𝑠) = {𝑌 (𝑠) 𝐶 (𝑠) − 𝑋 (𝑠)} {𝐼 + 𝐺 (𝑠) 𝐶 (𝑠)}−1 𝐷 ̃ (𝑠)] 𝑄𝑓 (𝑠) ,
𝑄 (𝑠) = [𝑄0𝑑 (𝑠) 𝑌 (𝑠) 𝑄 (24)
̃ (𝑠) .
− 𝑋 (𝑠) {𝐼 + 𝐺 (𝑠) 𝐶 (𝑠)}−1 𝐺 (𝑠) 𝑋 Then, from Figure 1, the state-space description of 𝐶𝑟 (𝑠) can
be obtained as
Because 𝐶(𝑠) stabilizes the closed system, 𝑄(𝑠) ∈ 𝐻∞ can 𝑥𝑟̇ (𝑡) = 𝐴 𝑟0 𝑥𝑟 (𝑡) + 𝐴 𝑟1 𝑥𝑟 (𝑡 − 𝐿) + 𝐵𝑟 𝑒 (𝑡) ,
be proven according to internal stability [15]. Therefore, the (27)
necessity is obtained. 𝑧 (𝑡) = 𝐶𝑟 𝑥𝑟 (𝑡 − 𝐿) ,
Sufficiency. Suppose that there is 𝑄(𝑠) ∈ 𝐻∞ written by (11) where 𝑥𝑟 (𝑡) is the state of 𝐶𝑟 (𝑠),
such that (10b) stabilizes the closed-loop system in Figure 1.
Substituting (10b) into (9) yields 𝐴𝑙
[ .. ]
𝐴 𝑟0 := [ . ],
̃ (𝑠) 𝑄0𝑑 (𝑠) + 𝐷 (𝑠) 𝑄0𝑛 (𝑠)}
𝐶1 (𝑠) = {𝑋
[ 𝐴 𝑙]
(22)
−1
̃ (𝑠) 𝑄0𝑑 (𝑠) − 𝑁 (𝑠) 𝑄0𝑛 (𝑠)} ,
× {𝑌 𝐵𝑙 𝑊1 𝐶𝑙 ⋅ ⋅ ⋅ 𝐵𝑙 𝑊𝑁−1 𝐶𝑙 𝐵𝑙 𝑊𝑁𝐶𝑙
[ 𝐵𝑙 𝐶𝑙 0 ] (28)
−1 [ ]
̃ (𝑠) {𝑌
𝐶2 (𝑠) = 𝑄 ̃ (𝑠) 𝑄0𝑑 (𝑠) − 𝑁 (𝑠) 𝑄0𝑛 (𝑠)} , (23) 𝐴 𝑟1 := [ . .. ] ,
[ .. . ]
[ 𝐵 𝐶
𝑙 𝑙 0 ]
respectively. Therefore, the sufficiency is obtained.
This completes the proof. 𝑇
𝐵𝑟 = [𝐵𝑙𝑇 0 ⋅ ⋅ ⋅ 0] , 𝐶𝑟 = [𝑊1 𝐶𝑙 ⋅ ⋅ ⋅ 𝑊𝑁𝐶𝑙 ] .
̃
From Theorem 1, 𝑄(𝑠) ∈ 𝑅𝐻∞ , 𝑄0𝑑 (𝑠) ∈ 𝑅𝐻∞ , and
̃
Assume that the state-space descriptions of 𝑄0𝑑 (𝑠), 𝑄(𝑠), and
𝑄0𝑛 (𝑠) ∈ 𝑅𝐻∞ are the free parameters that need to be deter-
̃ 𝑄0𝑑 (𝑠), ̃ are
𝑁(𝑠)
mined. In the following, the design methods for 𝑄(𝑠),
and 𝑄0𝑛 (𝑠) are based on BMIs and the control characteristic. ̃𝑞 𝐵̃𝑞
𝐴
𝐴 0𝑑 𝐵0𝑑
𝑄0𝑑 (𝑠) = [ ], ̃ (𝑠) = [
𝑄 ],
̃ 𝐶0𝑑 𝐷0𝑑 ̃𝑞 𝐷
𝐶 ̃𝑞
3.2. Design of Free Parameters 𝑄0𝑑 (𝑠) and 𝑄(𝑠). In [12, 13], the
designs of the free parameters use the Nyquist stability (29)
̃𝑛 𝐵̃𝑛
𝐴
criterion by manual examination, which is very difficult ̃ (𝑠) = [
𝑁 ],
and inefficient. To resolve this problem, it was essential to ̃𝑛 𝐷
𝐶 ̃𝑛
establish an efficient and easy design method for the free
parameters. In this subsection, an efficient design method where 𝐷0𝑑 has full column rank. Draw the configuration
for the free parameters using BMIs that can be solved with of 𝑄𝑓 (𝑠) in Figure 2 and the state-space description of the
existing software is presented. closed-loop system 𝑄𝑓 (𝑠) with 𝑟(𝑡) ≡ 0 is given by
The first step is to design the free parameters 𝑄0𝑑 (𝑠) and
̃
𝑄(𝑠) such that 𝑄(𝑠) belongs to 𝐻∞ on the basis of state-space 𝑥̇ (𝑡) = 𝐴 0 𝑥 (𝑡) + 𝐴 1 𝑥 (𝑡 − 𝐿) , (30)
Mathematical Problems in Engineering 5
𝐴 𝑟1
(𝑡) 𝑥𝑟 (𝑡) 𝑦𝑟 (𝑡)
𝑢(𝑡) + + 1
𝐵𝑟 𝐼 𝑒−𝑠𝐿 𝐶𝑟
+ 𝑠
𝐴 𝑟0
𝐶𝑟 (𝑠)
𝑟(𝑡) + 𝑢(𝑡)
− ̃𝑞
𝐷
𝐷0𝑑 − 𝐼
𝑦0 (𝑡) + 𝑥0 (𝑡) 1 𝑢(𝑡)
𝐶0𝑑 𝐼 + + 𝐵0𝑑
+ 𝑠
𝐴 0𝑑 ̃ 0𝑑 (𝑠)
𝑄
necessary to provide the stability condition of 𝑄0𝑑 (𝑠), and the and 𝑋 > 0 and appropriately dimensioned matrices Δ𝐴 ̃𝑞 , Δ𝐵̃𝑞 ,
following theorem will provide the stability condition in the Δ𝐶̃𝑞 , Δ𝐷
̃𝑞 , Δ𝐴0𝑑 , Δ𝐵0𝑑 , Δ𝐶0𝑑 , and Δ𝐷0𝑑 exist such that
form of BMIs for 𝑄0𝑑 (𝑠).
𝑇 𝑇 𝑇 𝑇
𝑋𝐴 + 𝐴 𝑋 𝑋𝐵0𝑑 𝐶0𝑑 Λ 𝑋 (𝐵0𝑑 + Δ𝐵0𝑑 ) 𝐶0𝑑 + Δ𝐶0𝑑
[ 0𝑑 ∗ 0𝑑 ] [∗ −𝐼 0 ]
Ξ := [ −𝐼 0 ]<0 (39) [ ]<0
𝑇 𝑇
[ ∗ ∗ −𝐷0𝑑 𝐷0𝑑 ] [ ∗ ∗ − (𝐷 0𝑑 + Δ𝐷 0𝑑 ) (𝐷 0𝑑 + Δ𝐷 )
0𝑑 ]
(44)
hold, then the time-delay system (30) is stable and 𝑄0𝑑 (𝑠) ∈
𝑅𝐻∞ . hold, where Λ, Δ𝐴 0 , and 𝐴 1 are defined as
Proof. From Lyapunov functional theory, 𝑄0𝑑 (𝑠) is a 𝑅𝐻∞ 𝑇 𝑇
matrix function if there exists a symmetric positive-definite Λ := 𝑋 (𝐴0𝑑 + Δ𝐴0𝑑 ) + (𝐴0𝑑 + Δ𝐴0𝑑 ) 𝑋, (45)
matrix 𝑋 = 𝑋𝑇 > 0 such that
̃𝑛
0 −𝐵𝑟 Δ𝐷0𝑑 𝐶 0 −𝐵𝑟 Δ𝐶0𝑑
[0 0 ̃
𝐵 Δ ̃
𝐶 0 ]
𝑋𝐴 0𝑑 + 𝐴𝑇0𝑑 𝑋 < 0, (40) [
Δ𝐴 0 := [ 𝑛 𝑞 ]
], (46)
[0 0 Δ𝐴̃𝑞 0 ]
̃
which is equivalent to [0 Δ𝐵0𝑑 𝐶𝑛 0 Δ𝐴0𝑑 ]
𝑇 𝑇 0 0 0 0
𝑋𝐴0𝑑 − 𝑋𝐵0𝑑 𝐷0𝑑 𝐶0𝑑 − (𝐵0𝑑 𝐷0𝑑 𝐶0𝑑 ) 𝑋 + 𝐴0𝑑 𝑋 < 0. (41) ̃𝑞 𝐶𝑟1
[𝐵̃𝑛 Δ𝐷 0 0 0]
Δ𝐴 1 := [
[ Δ𝐵̃ 𝐶
], (47)
𝑞 𝑟1 0 0 0]
According to the results in [16], a sufficient condition guar-
[ 0 0 0 0]
anteeing (41) is that there exists a positive number 𝜆 > 0 such
that
then the time-delay system (30) is stable and 𝑄0𝑑 (𝑠) is 𝑅𝐻∞ .
𝑇 𝑇
𝜆𝑋𝐴0𝑑 + 𝜆𝐴𝑇0𝑑 𝑋 +𝜆 2
𝑋𝐵0𝑑 𝐵0𝑑 𝑋 + 𝑇
𝐶0𝑑 𝐷0𝑑 𝐷0𝑑 𝐶0𝑑 < 0. ̃𝑞 , 𝐵̃𝑞 , 𝐶
̃𝑞 , 𝐷
̃𝑞 , 𝐴0𝑑 , 𝐵0𝑑 , 𝐶0𝑑 , and
Proof. Replacing matrices 𝐴
(42)
̃ ̃ ̃ ̃
𝐷0𝑑 with matrices 𝐴 𝑞 + Δ𝐴 𝑞 , 𝐵𝑞 + Δ𝐵𝑞 , 𝐶 ̃𝑞 + Δ𝐶 ̃𝑞 , 𝐷
̃𝑞 + Δ𝐷 ̃𝑞 ,
𝑇 𝐴0𝑑 + Δ𝐴0𝑑 , 𝐵0𝑑 + Δ𝐵0𝑑 , 𝐶0𝑑 + Δ𝐶0𝑑 , and 𝐷0𝑑 + Δ𝐷0𝑑 ,
Replacing 𝜆𝑋 and 𝐷0𝑑 𝑇
𝐷0𝑑 with 𝑋 and (𝐷0𝑑 𝐷0𝑑 )−1 and respectively, in (35) and (39), the constraints (44) can be
applying the Schur complement lemma show that (42) is obtained.
equivalent to (39). This completes this proof.
This completes the proof.
To avoid zero solutions, the initial transfer matrices for Constraints (44) are the BMIs that can be solved by BMI
̃ and 𝑄† (𝑠) are required. Without loss of generality, 𝑄(𝑠)
𝑄(𝑠) ̃ solvers, such as PENBMI [17] and bmibnb in YALMIP [18].
0𝑑
†
and 𝑄0𝑑 (𝑠) should be represented as
3.3. Design of Free Parameter 𝑄0𝑛 (𝑠). As mentioned, the
𝐴 ̃𝑞 𝐵̃𝑞 + Δ𝐵̃𝑞
̃𝑞 + Δ𝐴 objective of this paper is to develop a new design method
̃ (𝑠) = [
𝑄 ], that results in a closed-loop system in Figure 1 that is robustly
̃𝑞 + Δ𝐶
𝐶 ̃𝑞 + Δ𝐷
̃𝑞 𝐷 ̃𝑞 stable and has high control precision for reference inputs
(43) with uncertain period-time. Hence, the free parameter 𝑄0𝑛 (𝑠)
†
𝐴0𝑑 + Δ𝐴0𝑑 𝐵0𝑑 + Δ𝐵0𝑑 should be designed after the control characteristic. The output
𝑄0𝑑 (𝑠) = [ ],
𝐶0𝑑 + Δ𝐶0𝑑 𝐷0𝑑 + Δ𝐷0𝑑 sensitivity matrix function is defined as the transfer matrix
function from the reference input 𝑟(𝑡) to the tracking error
̃𝑞 , 𝐵̃𝑞 , 𝐶
̃𝑞 , 𝐷
̃𝑞 , 𝐴0𝑑 , 𝐵0𝑑 , 𝐶0𝑑 , and 𝐷0𝑑 are the given 𝑒(𝑡):
where 𝐴
initial matrices, and Δ𝐴 ̃𝑞 , Δ𝐵̃𝑞 , Δ𝐶 ̃𝑞 , Δ𝐷
̃𝑞 , Δ𝐴0𝑑 , Δ𝐵0𝑑 ,
𝑆 (𝑠) := (𝐼 + 𝐺 (𝑠) 𝐶 (𝑠))−1 . (48)
Δ𝐶0𝑑 , and Δ𝐷0𝑑 are the unknown matrices that need to
be determined. In the following theorem, modified stability Having good tracking performance of the closed-loop system
conditions are proposed in the form of BMIs. requires that
Theorem 4. For given initial matrix functions 𝑄(𝑠) ̃ and
†
𝑄0𝑑 (𝑠), if symmetric positive-definite matrices 𝑃 > 0, 𝑄 > 0, 𝜎 {𝑆 (𝑠)} = 𝜎 {(𝐼 + 𝐺 (𝑠) 𝐶 (𝑠))−1 } (49)
Mathematical Problems in Engineering 7
be as small as possible [15]. From (12), (10b), and (11), the For this reason, 𝑄0𝑛 (𝑠) and 𝑞(𝑠) are chosen by
sensitivity matrix function is written as
̃ (𝑠) 𝑄0𝑑 (𝑠) ,
𝑄0𝑛 (𝑠) = 𝑁𝑜+ (𝑠) 𝑞 (𝑠) 𝑌 (58)
̃ (𝑠) 𝑄0𝑑 (𝑠) − 𝑁 (𝑠) 𝑄0𝑛 (𝑠)}
𝑆 (𝑠) = {𝑌
{ 1 1 }
̃ (𝑠) − 𝑄0𝑑 (𝑠)) 𝐻 (𝑠)}−1
̃ (𝑠) 𝑄
× {𝑄0𝑑 (𝑠) + (𝑁 (50) 𝑞 (𝑠) = diag { 𝛼1 , . . . , 𝛼𝑝 } ∈ 𝑅𝐻∞ , (59)
(1 + 𝜏1 𝑠) (1 + 𝜏𝑝 𝑠)
{ }
̃ (𝑠) ,
× (𝐼 − 𝐻 (𝑠)) 𝐷
where 𝛼𝑖 (𝑖 = 1, . . . , 𝑝) is a positive integer to make 𝑄0𝑛 (𝑠)
and the largest singular value of the sensitivity matrix func- proper; then,
tion is
̃ (𝑠) 𝑄0𝑑 (𝑠) − 𝑁 (𝑠) 𝑄0𝑛 (𝑠) = (𝐼 − 𝑁𝑖 (𝑠) 𝑞 (𝑠)) 𝑌
𝑌 ̃ (𝑠) 𝑄0𝑑 (𝑠) .
̃ (𝑠) 𝑄0𝑑 (𝑠) − 𝑁 (𝑠) 𝑄0𝑛 (𝑠)}
𝜎 {𝑆 (𝑠)} ≤ 𝜎 {𝑌 (60)
Im
0
ble for both SISO linear systems and multiple-input/multiple-
output (MIMO) linear systems by simply modifying the −0.05
dimensions of some matrices.
−0.1
2
For this unstable control plant, it is easy to verify that the ̃ (𝑠) = 𝑠 + 5𝑠 + 6 , −1 𝑠2 + 5𝑠 + 16
𝑄 𝑄0𝑑 (𝑠) = . (70)
pairs (𝐴 𝑝 , 𝐵𝑝 ) and (𝐴 𝑝 , 𝐶𝑝 ) are controllable and observable, 𝑠2 + 15𝑠 + 50 𝑠2 + 5𝑠 + 6
respectively. According to [19], choose the 𝐹 and 𝐿 as follows:
Then, the feasible problem is solved using the PENBMI
𝐹 = [−3 −4] , 𝐿 = [−1 −4] .
𝑇
(64) ̃ and 𝑄0𝑑 (𝑠) as
solver to obtain 𝑄(𝑠)
0 5. Conclusion
Magnitude (dB)
Acknowledgments
period-time had a limited uncertainty. When the reference
had intermediate frequency components, the control perfor- The first author would like to express his gratitude to the
mance became worse than that without the repetitive con- China Scholarship Council (CSC) for its support and schol-
troller. However, the control performance at the intermediate arship. The authors gratefully acknowledge the reviewers
frequencies was improved by modifying the cutoff frequency for their helpful comments and suggestions, which have
of the low-pass filter 𝑞(𝑠). improved the presentation.
The simulation results using our method are shown in
Figure 5. Clearly, the high-order modified repetitive control
system was stable. For the period-time without variation, the
References
largest steady-state peak-to-peak relative error was as small [1] S. Hara, Y. Yamamoto, T. Omata, and M. Nakano, “Repetitive
as 0.05%, and this system entered the steady state in the control system: a new type servo system for periodic exogenous
11th period. When the period-time had an uncertainty level signals,” IEEE Transactions on Automatic Control, vol. 33, no. 7,
Δ of 5%, the largest steady-state peak-to-peak relative error pp. 659–668, 1988.
was 0.15%. This result shows that for a certain vibration in [2] B. A. Francis and W. M. Wonham, “The internal model principle
period-time, the high-order modified repetitive controller for linear multivariable regulators,” Applied Mathematics &
continued to provide high control precision. As seen in Optimization, vol. 2, no. 2, pp. 170–194, 1975.
Figure 4, when Δ = 11%, the largest steady-state peak-to- [3] B. Mahawan and Z. H. Luo, “Repetitive control of tracking
peak relative error of this control system increases to 1.6%. systems with time-varying periodic references,” International
This tracking error is small and can be further reduced by Journal of Control, vol. 73, no. 1, pp. 1–10, 2000.
modifying the cutoff frequency of the low-pass filter 𝑞(𝑠) [4] J. S. Freudenberg and D. P. Looze, “Right half plane poles and
because it has no bandwidth restriction. The results shown zeros and design tradeoffs in feedback systems,” IEEE Transac-
in Figure 5 demonstrate that the design method presented tions on Automatic Control, vol. 30, no. 6, pp. 555–565, 1985.
here provided not only good stability, but also satisfactory [5] G. Pipeleers, B. Demeulenaere, J. De Schutter, and J. Swevers,
tracking performance for the reference inputs with uncertain “Robust high-order repetitive control: optimal performance
period-time. Compared with conventional design methods, trade-offs,” Automatica, vol. 44, no. 10, pp. 2628–2634, 2008.
the design method proposed in this paper widens the range [6] T. Inoue, “Practical repetitive control system design,” in Pro-
of the uncertainty in period-time and has a wider range of ceedings of the 29th IEEE Conference on Decision and Control,
practical applications, which makes it very practical. pp. 1673–1678, Honolulu, Hawaii, USA, December 1990.
10 Mathematical Problems in Engineering
International
Journal of Journal of
Mathematics and
Mathematical
Discrete Mathematics
Sciences