research article generalized kalman-yakubovich-popov...
TRANSCRIPT
Hindawi Publishing CorporationJournal of Applied MathematicsVolume 2013 Article ID 854631 9 pageshttpdxdoiorg1011552013854631
Research ArticleGeneralized Kalman-Yakubovich-Popov Lemma Based I-PDController Design for Ball and Plate System
Shuichi Mochizuki1 and Hiroyuki Ichihara2
1 Graduate School of Science and Technology Meiji University 1-1-1 Higashimita Tama-ku Kawasaki-shi Kanagawa 214-8571 Japan2Department of Science and Engineering Meiji University 1-1-1 Higashimita Tama-ku Kawasaki-shi Kanagawa 214-8571 Japan
Correspondence should be addressed to Hiroyuki Ichihara ichiharamessemeijiacjp
Received 23 June 2013 Accepted 9 September 2013
Academic Editor Baocang Ding
Copyright copy 2013 S Mochizuki and H IchiharaThis is an open access article distributed under theCreative CommonsAttributionLicense which permits unrestricted use distribution and reproduction in anymedium provided the originalwork is properly cited
The ball-on-plate balancing system has a camera that captures the ball position and a plate whose inclination angles are limitedThispaper proposes a PID controller design method for the ball and plate system based on the generalized Kalman-Yakubovich-Popovlemma The design method has two features first the structure of the controller called I-PD prevents large input signals againstmajor changes in the reference signal second a low-pass filter is introduced into the feedback loop to reduce the influence of themeasurement noise produced by the camera Both simulations and experiments are used to evaluate the effectiveness of the designmethod
1 Introduction
The ball and plate [1] is an unstable underactuated nonlinearsystem that has double integrators at the origin and that hastwo control inputs against four degrees of freedom (DOF) Acamera located above the plate captures the position of theball and two motors manipulate the inclination angles of theplate to keep the ball on the plateThe ball and plate system isan extension of the ball and beam system [2] from one to twodimensions The system has demonstrated various controllerdesignmethods for positioning and trajectory tracking of theball proportional integral derivative (PID) control [3] fuzzycontrol [4] neural network control [3 5] and model predic-tive control [6] In particular PID control has the benefitsof simple implementation and fewer hardware requirementsand it has been applied in many successful designs [7]Because PID control enables us a limited performanceoptimizing the parameters in a PID controller satisfyingdesign specifications is an important subject for study In thecontroller design of the ball and plate it requires to considerlimitation of the inclination angles with good transient andsteady-state responses Although proportional and derivativecontrollers are required to improve transient responses ajump in the reference signal generates a large input signal
that reaches the limitation angle that degrades the transientresponses [8] In addition ball position data from the camerainclude measurement noise that also degrades the steady-state responses
To overcome the above issues this paper proposes aPID controller design method for the ball and plate systemby open-loop shaping based on the generalized Kalman-Yakubovich-Popov (GKYP) lemma [9]TheGKYP lemma is ageneralization of the standard KYP lemma [10] which estab-lishes the equivalence between a frequency domain inequality(FDI) for a transfer function and a linear matrix inequality(LMI) associatedwith its state-space realizationThe standardKYP lemma is available for the infinite frequency rangewhile the generalized one can limit the frequency range tobe (semi) finite By introducing the GKYP lemma to PIDcontroller design design specifications by FDIs in the finitefrequency ranges for the open-loop transfer function resultin LMIs [9] The GKYP lemma gives a systematic open-loop shaping design method through optimization to realizedesirable transient and steady-state responses In this visualfeedback system we introduce a low-pass filter Since thefilter gives freedom in optimization it allows better steady-state responses and reduces the influence of themeasurementnoise To prevent large input signals fromP andD controllers
2 Journal of Applied Mathematics
Figure 1 2D Ball Balancer
we adopt the I-PD (integral-proportional derivative) struc-ture whose design is still in the framework of the GKYPlemma because the open-loop transfer functions of PID andI-PD structures are fundamentally the same
The paper is organized as follows The description of theball and plate system including its modeling and the mea-surement noise is presented in Section 2 The GKYP lemmabased I-PD controller design method with a low-pass filteris provided in Section 3 The design of the I-PD controller isdescribed in Section 4 Simulation and experimental resultsare presented in Sections 5 and 6 respectively Finally inSection 7 we present our conclusions
The notation used is standard For a matrix 119872 thetranspose and complex conjugate transpose are denoted by119872⊤ and 119872lowast respectively For a Hermitian matrix 119872 119872 ≻
(⪰) 0 and 119872 ≺ (⪯) 0 denote positive (semi) definitenessand negative (semi) definiteness respectivelyThe symbolH
119899
stands for the set of 119899 times 119899 Hermitian matrices The subscript119899 is omitted if 119899 = 2 The real and imaginary parts of 119872are denoted by R(119872) and I(119872) For matrices Φ and 119875Φotimes119875 denotes the Kronecker productL119909(119905) represents theLaplace transform of a signal 119909(119905)
2 Ball and Plate System
The ball and plate a QUANSER 2D Ball Balancer is shownin Figure 1 The system consists of a plate a ball an overheadcamera and two servo units The plate is allowed to swivelin both the 119883- and 119884-directions The overhead CMOSdigital camera a Point Grey Research Inc FFMV-03M2C-CSmeasures the position of the ball The two servo units locatedunder the plate are QUANSER SRV02 devices each of whichhas a peak time of approximately 200 ([ms]) and an overshootof approximately 5 Each of the devices is connected to aside of the plate through a two DOF gimbal The samplingtime of the control system and the frame rate provided bythe camera are 1 ([ms]) and 60 ([fps]) respectively Thus the
Motorgear
Loadgear
Potentiometergear
Ball
Plate
Supportbeam Bottom
support plate
r arm
120572
120579x
(xb yb)
x
Ltbl
0
Figure 2 The ball and plate system
image information is renewed approximately every 17 ([ms])There is a constant time delay of less than 60 ([ms]) betweenthemeasurement of the ball position and themanipulation ofthe servo units in the visual feedback system
21 Modeling The119883-direction of the ball and plate system isillustrated in Figure 2 We assume that the ball is completelysymmetric and homogeneous and does not slip on the plateand that all frictions are neglected The plate rotates in the119883119884-Cartesian coordinates with the origin at the center of theplate The equations of motion are
(119898119887+119868119887
1199032
119887
) 119887minus 119898119887(1199091198872+ 119910119887 120573) + 119898
119887119892 sin120572 = 0
(119898119887+119868119887
1199032
119887
) 119910119887minus 119898119887(119910119887
1205732+ 119909119887 120573) + 119898
119887119892 sin120573 = 0
(1)
where (119909119887 119910119887) is the position of the ball on the plate 120572 and
120573 are the inclination angles of the plate to the 119883- and 119884-axisrespectively 119898
119887is the mass of the ball 119903
119887is the radius of the
ball 119892 is the gravitational acceleration and 119868119887is the inertia of
the ball In Figure 2 120579119909represents the angle of the load gear
The relationship between 120572 and 120579119909is as follows
sin120572 =2 sin 120579
119909119903arm
119871 tbl (2)
where 119871 tbl is the length of the side of the plate and 119903arm is thelength between the joint and the center of the load gear Therelationship of 120573 and 120579
119910is the same as (2) since both gear
systems have the same hardware and the plate is symmetricalThe numerical values of the constant parameters in theequations of motion and (2) are shown in Table 1 Since 120579
119909
and 120579119910are limited as
minus30 [∘] le 120579
119909 120579119910 le 30 [
∘] (3)
from (2) the working ranges of 120572 and 120573 are
minus53 [∘] le 120572 120573 le 53 [
∘] (4)
If the angular velocities and 120573 are relatively low theapproximations
120573 = 0 2= 0 120573
2= 0 (5)
Journal of Applied Mathematics 3
Table 1 Parameters of the ball and plate system
Parameters Numerical values119898119887
00252 [kg]119903119887
00170 [m]119892 981 [ms2]119868119887
289 times 10minus6 [kgsdotm]
119871 tbl 0275 [m]119903arm 00254 [m]
are often used Linearizing the equations of motion at 120579119909= 0
and 120579119910= 0 we have
(119898119887+119868119887
1199032
119887
) 119887+2119898119887119892119903arm119871 tbl
120579119909= 0
(119898119887+119868119887
1199032
119887
) 119910119887+2119898119887119892119903arm119871 tbl
120579119910= 0
(6)
Since the axes are independent of each other we can focuson one axis for example the119883-axis For the input 120579
119909and the
output 119909119887 the transfer function is given by
119875 (119904) =119883119887(119904)
Θ119909(119904)
=
119870bap
1199042 (7)
where119883119887(119904) = L119909
119887(119905) Θ
119909(119904) = L120579
119909(119905) and
119870bap = minus2119898119887119892119903arm119903
2
119887
119871 tbl (1198981198871199032
119887+ 119868119887) (8)
22 Measurement Noise In this visual feedback system thereis inevitable noise from the camera To examine the noiselevel and frequencies we observed the error signal betweena fixed ball position and a measurement signal The resultsare shown in Figure 3 where the upper part represents a timehistory of the error signal including noise and the lower partrepresents the fast Fourier transform (FFT) analysis of theerror signal The noise level in the error signals is relativelyhigh at frequencies over 20 ([rads])
3 I-PD Control by GKYP Lemma
This section describes an I-PD controller design methodbased on the GKYP lemma The feedback control system isshown in Figure 4 where a filter is introduced into the controlsystem
31 Low-Pass Filter In the previous section we showed thatthemeasurement noise degrades the control performance Toreduce the influence of the noise a low-pass filter is availablein the controller design According to the noise propertiesthat we observed it is sufficient to introduce a first-order low-pass filter into the output of the measurement such that
119865 (119904) =120583
119904 + 120583 (9)
where 120583 ([rads]) is the cut-off frequency
0 1 2 3 4 5 6 7 8Time (s)
Am
plitu
de (m
)
002040608
112141618
2
Frequency (rads)
Pow
er
minus49
minus485
minus48
minus475times10minus3
100 101 102
Figure 3 Time history and FFT analysis results of measumentsignal
r e
minus minus
I-PD controlleru
Actuator Ball and platesystem
Low-passfilter
I
P + D
Figure 4 Feedback control system of the ball and plate
32 I-PD Controller In the standard PID control majorchanges in the reference signals cause large input signals tobe generated by the proportional and derivative actions in thecontroller that saturate the actuator Indeed it is difficult totune the parameters in the PID controller (Figure 4) such thatthe actuator in this visual feedback system is not saturatedThe control system with the I-PD controller (Figure 4) hasa structure whose inner loop includes the proportional andderivative actions [7 8] In this structure the integral actionalone acts on the error signal and prevents large signals beinginput to the actuator The control input 119906 can be written as
119906 = minus119870119901119910 +
119870119894
119904(119903 minus 119910) minus
119870119889119904
120591119904 + 1119910 (10)
where 120591(gt 0) is the parameter to approximate the differen-tiator by a proper transfer function119870
119901119870119894 and119870
119889represent
the proportional integral and derivative gains respectivelyThe open-loop transfer function is
119871 (119904) = 119865 (119904) 119875 (119904)119870 (119904) (11)
where
119870 (119904) = 119870119901+119870119894
119904+
119870119889119904
120591119904 + 1= 119870119901(1 +
1
119879119894119904+
119879119889119904
120591119904 + 1) (12)
119879119894and 119879
119889are the integral time and derivative time respec-
tively It should be noted that the open-loop transfer function
4 Journal of Applied Mathematics
of the I-PD structure is the same as that of the standardPID structure To tune the parameters in the I-PD controllerwe employ an open-loop shaping that realizes a desirablefrequency response of the closed-loop system
33 Generalized KYP Lemma It is known that design spec-ifications for an open-loop transfer function can be reducedto LMIs through the GKYP lemma [9] We briefly review thislemma in the case of continuous-time systems
The design specification consists of a frequency range anda desired property in that range The frequency range can berepresented by
Λ (ΦΨ) = 119904 isin C | 120590 (119904 Φ) = 0 120590 (119904 Ψ) ge 0 (13)
where ΦΨ isin H
120590 (119904 Φ) = [119904lowast1]Φ[
119904
1] = 0 (14)
The equality constraint in (13) distinguishes betweencontinuous-time and discrete-time specifications Since weaddress continuous-time systems in this paper we use Φsuch that
Φ = [0 1
1 0] (15)
The inequality constraint 120590(119904 Ψ) ge 0 in (13) sets a frequencyrangeΩ For example a low frequency range is written as
Ω = 120596 | 120596 le 120603119897 = 120596 | 120590 (119895120596 Ψ) ge 0 (16)
where
Ψ = [0 minus119895
119895 2120603119897
] (17)
Table 2 presents a summary of the choice of Ψ versus a typeof the frequency range Ω where 120603
ℓ 120603ℎ 1206031 and 120603
2are real
positive numbers and 120603119888= (1206031+ 1206032)2 On the other hand
the desired property in a specific frequency range can berepresented by
[119871 (119895120596) 119868]Π [119871lowast(119895120596)
119868] ≺ 0 (18)
where
Π = [Π11
Π12
Π21
Π22
] isin H119898+119901
Π11isin H119901 Π ⪰ 0 (19)
119898 and 119901 are the input and output numbers of 119871(119904) respec-tively For SISO systems consider the requirement that 119871(119895120596)in a frequency range is on the half plane under a straight lineThat is 119871(119895120596) is under the straight line such that
119886R [119871 (119895120596)] + 119887I [119871 (119895120596)] lt 119888 (20)
that is equivalent to (18) with
Π = [0 119886 minus 119895119887
119886 + 119895119887 minus2119888] isin H
2 (21)
Table 2 Relation between Ψ andΩ for continuous time
Ψ Ω
[0 minus119895
119895 2120603ℓ
] 120596 le 120603ℓ(low)
[minus1 119895120603
119888
minus11989512060311988812060311206032
] 1206031le 120596 le 120603
2(middle)
[0 119895
minus119895 minus2120603ℎ
] 120603ℎle 120596 (high)
This requirement is designed to reduce sensitivity in a lowfrequency range Another requirement is that 119871(119895120596) in afrequency range is in the interior of the circle of radius 119903withthe center at 119888 That is 119871(119895120596) is in the circle such that
1003816100381610038161003816119871 (119895120596) minus 1198881003816100381610038161003816
2
lt 1199032 (22)
which is equivalent to (18) with
Π = [1 minus119888
lowast
minus119888 |119888|2minus 1199032] isin H
2 (23)
This requirement is designed to guarantee robustness in ahigh frequency range Under these preparations the gener-alized KYP lemma [9] is expressed as follows
Lemma 1 Let 119871(119904) be 119862(119904119868 minus 119860)minus1119861 + 119863 Λ(ΦΨ) in (13) andΠ in (18) are given Assume that det(119904119868 minus 119860) = 0 for all 119904 isin ΛThen the finite frequency condition
[119871 (119904) 119868]Π [119871lowast(119904)
119868] ≺ 0 forall119904 isin Λ (ΦΨ) (24)
holds if and only if there exist Hermitianmatrices119875 and119876 suchthat the matrix inequality condition
[[[
[
Γ [119861
119863]Π11
Π11[119861
119863]
lowast
minusΠ11
]]]
]
≺ 0 (25)
is satisfied where
Γ = [119860 119868
119862 0] (Φ⊤otimes 119875 + Ψ
⊤otimes 119876) [
119860 119868
119862 0]
⊤
+ [
[
0 119861Π12
Πlowast
12119861⊤119863Π12+ Πlowast
12119863⊤+ Π22
]
]
(26)
Equation (25) is affine with respect to 119861119863 119875119876 andΠ22
In the case where 119861 and119863 have affine design parameters (25)is an LMI
4 Control System Design
41 Filter Design Considering the control performance andthe noise level we set the cut-off frequency in (9) as 120583 =
20 ([rads]) We examine the effectiveness of this filter in
Journal of Applied Mathematics 5
LPF
0 1 2 3 4 5 6 7 8Time (s)
Am
plitu
de (m
)
002040608
112141618
2
Frequency (rads)
Pow
er
minus49
minus485
minus48
minus475
times10minus3
100 101 102
Raw data
Figure 5 Time history and FFT analysis result of measurementsignal with low-pass filter
the same experimental setup as given in Section 22 Figure 5shows the spectral analysis results of the measurement andfiltered signals which are represented by the dotted and solidcurves respectively From these results it can be seen that thenoise at frequencies over 20 ([rads]) has been reduced
42 State-Space Realization of Open-Loop Transfer FunctionTo obtain an LMI based on the GKYP lemma a state-spacerealization of 119871(119904) is required to be affine with respect to aset of the design parameters 120588 = (119870
119901 119870119894 119870119889) If we fix 120591 in
119870(119904) at 10times10minus2 the design parameters appear affinely in thenumerator of119870(119904) Indeed the controllable canonical form of119870(119904) is written as
[119860119896119861119896(120588)
119862119896119863119896(120588)
] =
[[[[[[
[
0 0119870119894
120591
1 minus1
120591119870119894minus119870119889
1205912
0 1 119870119901+119870119889
120591
]]]]]]
]
(27)
Realizations of 119875(119904) and 119865(119904) are also written as
[119860119901
119861119901
119862119901119863119901
] = [
[
0 1 0
0 0 119870bap1 0 0
]
]
(28)
[119860119891
119861119891
119862119891
119863119891
] = [minus120583 1
120583 0] (29)
respectively By combining these realizations (27)ndash(29) weobtain a realization of 119871(119904) as
119871 (119904) = [119860 119861 (120588)
119862 119863 (120588)] (30)
where
119860 = [
[
119860119896
0 0
119862119896119861119901
119860119901
0
119862119896119863119901119861119891119862119901119861119891119860119891
]
]
119861 = [
[
119861119896(120588)
119863119896(120588) 119861119901
119863119896(120588)119863119901119861119891
]
]
119862 = [119862119896119863119901119861119891 119862119901119861119891119862119891] 119863 = 119863
119896(120588)119863119901119863119891
(31)
Consequently the state-space realization of 119871(119904) is affine withrespect to 120588
43 Specifications To shape the Nyquist plot of 119871(119904) werequire the following FDI specifications
minus2R [119871 (119895120596)] +I [119871 (119895120596)] gt 120574ℓforall120596 le 08 (32)
I [119871 (119895120596)] lt 120574119898 25 le
forall120596 le 28 (33)
1003816100381610038161003816119871 (119895120596)1003816100381610038161003816 lt 120574ℎ
forall120596 ge 10 (34)
Specification (32) with a large 120574ℓ(gt 0) ensures sensitivity
reduction in the low frequency range by making the gain of119871(119904) high Specification (33) requires the Nyquist plot to beoutside a circle with its center at the point minus1 + 1198950 so thata certain stability margin is guaranteed Specification (34)with a small 120574
ℎensures robustness against the unmodeled
dynamics that typically exists in the high frequency rangeIn addition to the above basic specifications we also
require the following FDIs that improve the property oftrajectory tracking
4R [119871 (119895120596)] +I [119871 (119895120596)] lt 1205741forall120596 ge 03 (35)
4R [119871 (119895120596)] +I [119871 (119895120596)] gt 1205742forall120596 ge 03 (36)
Since the integral action aloneworks on the error between theoutput and the reference signals the property of trajectorytracking depends directly on the integrator Here we focuson the corner angular frequency 120596
119868by the integral action
in the I-PD controller which is given by 120596119868= 1119879
119894 where
119879119894= 119870119901119870119894 We have a strong integral action and the error is
corrected quickly when the corner angular frequency is highwhile too high a corner angular frequency causes overshootand hunting Thus we impose restrictions for the phase ofthe I-PD controller It should be noted that the phase atlower frequencies is about minus90 ([deg]) while the phase atthe corner angular frequency is about 0 [deg] If we find thefrequency at a specific phase from minus90 ([deg]) to 0 ([deg])the corner angular frequency is greater than that frequencySpecifications (35) and (36) restrict the phase of the I-PDcontroller as well as the open-loop transfer function so thatthe corner angular frequency is greater than the frequency atthe lowest point in the frequency range 03 ([rads])
44 I-PD Controller Design We design an I-PD controller bymaximizing 120574
ℓsubject to Specifications (32)ndash(36) where
(120574119898 120574ℎ 1205741 1205742) = (minus1 05 1 minus100) (37)
6 Journal of Applied Mathematics
0 50 100 150 200
0
50
100
150
Real axis
Imag
inar
y ax
is
The upper limit
Low frequency
The lower limit
minus150
minus100
minus50
minus150minus200 minus100 minus50
Figure 6 Nyquist plot
That is the optimization problem is
max119870119894 119870119901119870119889 120574ℓ
120574ℓ
subject to (19)ndash(36) and (37)
(38)
Each of the design Specifications (32)ndash(36) is reduced to anLMI condition through Lemma 1 with the realization (30)The Specification (32) is modified to 120576 le forall120596 le 08 where 120576 =10 times 10
minus4 because 119871(119904) includes the origin poles that preventus from taking 120596 = 0 Then the LMI optimization problemis to maximize 120574
ℓsubject to all these LMI conditions where
119870119894 119870119901 and 119870
119889are the common decision variables while
1198751 119875
5and 119876
1 119876
5appear in the LMIs as independent
decision variables It should be noted that 120574ℓappears alone in
Π22in the LMI condition (25) corresponding to (36) In this
sense 120574ℓis also an independent decision variable It should
also be noted that 119875 and 119876 in (25) appear in each of the LMIconditions as the independent decision variables
To solve this LMI optimization problem we use YALMIPR20120806 [11] an LMI parser and SPDT3 version 40 [12]an LMI solver on MATLAB R2011b The resulting optimalparameters in the I-PD controller and 120574
ℓare
(119870119901 119870119894 119870119889 120574ℓ) = (92859 107806 4138 119464)
(39)
The Nyquist plots are shown in Figures 6 and 7 where 119871(119895120596)satisfies the design specifications given in Section 43 Since 120574
ℓ
is maximized sensitivity is reduced in the frequency rangeThe Bode plots of 119865(119904)119875(119904) 119871(119904) and 119870(119904) are shown inFigure 8 where the corner angular frequency 120596
119868of the I-PD
controller is larger than 03 ([rads])
5 Simulation Results
The I-PD controller whose design is described in the previoussection was evaluated by a simulation of the step response Tocompare the response with that of a standard PID controller
0 1
0
05
1
15
Real axis
Imag
inar
y ax
is
Middle frequency
High frequency
minus15
minus1
minus05
minus5 minus4 minus3 minus2 minus1
120596 = 102 (rads)
120596 = 30 (rads)
Figure 7 Nyquist plot (magnification)
0
100
200
Gai
n (d
B)
0
90
Phas
e (de
g)
Frequency (rads)
minus300
minus200
minus100
minus270
minus180
minus90
10minus2 100 102 104
P(s) + F(s)
L(s)
K(s)
Frequency (rads)10minus2 100 102 104
Figure 8 Bode plot
we used the PID controller whose gain parameters are thesame as those of the I-PD controller Since both feedbacksystems have the same open-loop transfer function theirfeedback properties must be the same provided that eachinput signal does not saturate The simulation results of thestep response are shown in Figure 9 where the upper andlower parts are the output and input signals respectivelyThe solid curves represent the responses given by the I-PDcontroller while the dashed curves represent those by the PIDcontroller One can see that the input signal given by the PIDcontroller is saturated while that by the I-PD controller is notsaturated and satisfies the limitation (3) The output signalgiven by the I-PD controller settles down to the desired valuewithout any overshoot
It should be noted that the gain parameters in thedesigned controller are not tuned with the I-PD structure
Journal of Applied Mathematics 7
0 1 2 3 4 50
005
01
015
Am
plitu
de (m
)
0 1 2 3 4 5
0102030
Time (s)
Ang
le (d
eg)
I-PDPID
minus30
minus20
minus10
Time (s)
Figure 9 Step response simulation results
0 05 1 15 2 25 3 35 4 45 50
005
01
015
Time (s)
Am
plitu
de (m
)
0 05 1 15 2 25 3 35 4 45 5
0102030
Am
plitu
de (r
ad)
minus30
minus20
minus10
Time (s)
I-PDPID
Figure 10 Step response experimental results
In our experience it is difficult not to saturate the inputlimitation for the PID structure using any design method
6 Experimental Results
This section evaluates the I-PD controller whose designis given in Section 4 through an experiment of the stepresponse The PID controller used in Section 4 was alsoevaluated for comparison The results of trajectory trackingcontrol by the I-PD controller were also evaluated
61 Step Response Experiment The results of the stepresponse experiment are shown in Figure 10 where thedescription of the figure is the same as that of Figure 9 In thisexperiment the influence of the time delay appeared and theinput signals were slightly larger than those obtained in thesimulations The rise and settling time results are howeveralmost the same as those obtained in the simulation
62 Trajectory Tracking Experiments We tested two kindsof trajectories for tracking control a square and a circular
8 Journal of Applied Mathematics
0 002 004
0
002
004
006
Reference
minus006minus006
minus004
minus004
minus002
minus002 0 002 004
0
002
004
006
minus006minus006
minus004
minus004
minus002
minus002
yb
(m)
xb (m)
I-PD controlyb
(m)
xb (m)
ReferenceI-PD control
Figure 11 Experimental results of tracking control
0 5 10 15 20 25 30 35 40 45 50
0005
01
02040
Time (s)
0 5 10 15 20 25 30 35 40 45 50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
minus01minus005
minus40minus20
02040
minus40minus20
0005
01
minus01minus005
Time (s)
Time (s)Time (s)
ReferenceI-PD control
Controller outputActuator output
xb
(m)
120572(d
eg)
yb
(m)
120573(d
eg)
Figure 12 Experimental results of square response
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0005
01
02040
minus01minus005
minus40minus20
02040
minus40minus20
0005
01
minus01minus005
Time (s)Time (s)
Time (s)Time (s)
xb
(m)
120572(d
eg)
yb
(m)
120573(d
eg)
ReferenceI-PD control
Controller outputActuator output
Figure 13 Experimental results of circle response
Journal of Applied Mathematics 9
trajectory The side of the square trajectory was 01 ([m])and the radius of the circular trajectory was 005 ([m]) Theresults are shown in Figure 11 where the left part shows atrajectory of the square trajectory tracking control and theright part shows a trajectory of the circular trajectory trackingcontrol The time histories of the ball and input angles areshown in Figures 12 and 13 In the square trajectory trackingcontrol experiment the responses were similar to those inthe step response experiment except for a slight vibrationSuch vibration phenomena are noticeable in the responsesof circular trajectory tracking control in particular the casewhen the input signal is relatively small The reason for thesephenomena could be the friction of the ball against the plateor a backlash of the gear system
7 Conclusions
This paper applied the GKYP lemma to an open-looptransfer function including an I-PD controller and a noisereduction filter for the ball and plate systemThemultiple FDIspecifications for the finite frequency ranges were satisfiedby a solution of the LMI optimization problem The solutionincludes the optimal parameters in the I-PD controller Thefirst-order low-pass filter reduced the noise in the high fre-quency range and improved the steady-state response Bothsimulations and experiments evaluated the effectiveness ofthe designmethod by comparing the standard PID controller
The PI-D (proportional integral-derivative) control sys-tem which moved the derivative controller to the innerfeedback loop also has the same open-loop transfer functionas the standard PID controller Thus the approach in thispaper can also be applied to the PI-D controller
References
[1] MMoarref M Saadat and G Vossoughi ldquoMechatronic designand position control of a novel ball and plate systemrdquo inProceedings of the Mediterranean Conference on Control andAutomation (MEDrsquo 08) pp 1071ndash1076 June 2008
[2] P Kokotovic ldquoThe joy of feedback nonlinear and adaptiverdquoIEEE Control Systems Magazine vol 12 no 3 pp 7ndash17 1992
[3] F Zheng X Li X Qian and S Wang ldquoModeling and PIDneural network research for the ball and plate systemrdquo inProceedings of the International Conference on Electronics Com-munications and Control (ICECC rsquo11) pp 331ndash334 September2011
[4] X Fan N Zhang and S Teng ldquoTrajectory planning andtracking of ball andplate systemusing hierarchical fuzzy controlschemerdquo Fuzzy Sets and Systems vol 144 no 2 pp 297ndash3122004
[5] M Bai Y Tian and Y Wang ldquoDecoupled fuzzy sliding modecontrol to ball and plate systemrdquo in Proceedings of the 2ndInternational Conference on Intelligent Control and InformationProcessing (ICICIP rsquo11) pp 685ndash690 July 2011
[6] F Borrelli Constrained Optimal Control of Linear and HybridSystems vol 290 of Lecture Notes in Control and InformationSciences Springer 2003
[7] K J Astrom and T Hagglund Advanced PID Control Instru-mentation Systems 2005
[8] Q Li and Z Kemin Introduction to Feedback Control PrenticeHall 2009
[9] T Iwasaki and S Hara ldquoGeneralized KYP lemma unifiedfrequency domain inequalities with design applicationsrdquo IEEETransactions on Automatic Control vol 50 no 1 pp 41ndash592005
[10] A Rantzer ldquoOn the Kalman-Yakubovich-Popov lemmardquo Sys-tems amp Control Letters vol 28 no 1 pp 7ndash10 1996
[11] J Lofberg ldquoYALMIP a toolbox for modeling and optimizationin MATLABrdquo in Proceedings of the IEEE International Sympo-sium on Computer-Aided Control System Design (CACSD rsquo11)Taipei Taiwan 2004
[12] K C TohM J Todd and RH Tutuncu ldquoSDPT3mdashaMATLABsoftware package for semidefinite programming version 13rdquoOptimization Methods and Software vol 11 no 1 pp 545ndash5811999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
2 Journal of Applied Mathematics
Figure 1 2D Ball Balancer
we adopt the I-PD (integral-proportional derivative) struc-ture whose design is still in the framework of the GKYPlemma because the open-loop transfer functions of PID andI-PD structures are fundamentally the same
The paper is organized as follows The description of theball and plate system including its modeling and the mea-surement noise is presented in Section 2 The GKYP lemmabased I-PD controller design method with a low-pass filteris provided in Section 3 The design of the I-PD controller isdescribed in Section 4 Simulation and experimental resultsare presented in Sections 5 and 6 respectively Finally inSection 7 we present our conclusions
The notation used is standard For a matrix 119872 thetranspose and complex conjugate transpose are denoted by119872⊤ and 119872lowast respectively For a Hermitian matrix 119872 119872 ≻
(⪰) 0 and 119872 ≺ (⪯) 0 denote positive (semi) definitenessand negative (semi) definiteness respectivelyThe symbolH
119899
stands for the set of 119899 times 119899 Hermitian matrices The subscript119899 is omitted if 119899 = 2 The real and imaginary parts of 119872are denoted by R(119872) and I(119872) For matrices Φ and 119875Φotimes119875 denotes the Kronecker productL119909(119905) represents theLaplace transform of a signal 119909(119905)
2 Ball and Plate System
The ball and plate a QUANSER 2D Ball Balancer is shownin Figure 1 The system consists of a plate a ball an overheadcamera and two servo units The plate is allowed to swivelin both the 119883- and 119884-directions The overhead CMOSdigital camera a Point Grey Research Inc FFMV-03M2C-CSmeasures the position of the ball The two servo units locatedunder the plate are QUANSER SRV02 devices each of whichhas a peak time of approximately 200 ([ms]) and an overshootof approximately 5 Each of the devices is connected to aside of the plate through a two DOF gimbal The samplingtime of the control system and the frame rate provided bythe camera are 1 ([ms]) and 60 ([fps]) respectively Thus the
Motorgear
Loadgear
Potentiometergear
Ball
Plate
Supportbeam Bottom
support plate
r arm
120572
120579x
(xb yb)
x
Ltbl
0
Figure 2 The ball and plate system
image information is renewed approximately every 17 ([ms])There is a constant time delay of less than 60 ([ms]) betweenthemeasurement of the ball position and themanipulation ofthe servo units in the visual feedback system
21 Modeling The119883-direction of the ball and plate system isillustrated in Figure 2 We assume that the ball is completelysymmetric and homogeneous and does not slip on the plateand that all frictions are neglected The plate rotates in the119883119884-Cartesian coordinates with the origin at the center of theplate The equations of motion are
(119898119887+119868119887
1199032
119887
) 119887minus 119898119887(1199091198872+ 119910119887 120573) + 119898
119887119892 sin120572 = 0
(119898119887+119868119887
1199032
119887
) 119910119887minus 119898119887(119910119887
1205732+ 119909119887 120573) + 119898
119887119892 sin120573 = 0
(1)
where (119909119887 119910119887) is the position of the ball on the plate 120572 and
120573 are the inclination angles of the plate to the 119883- and 119884-axisrespectively 119898
119887is the mass of the ball 119903
119887is the radius of the
ball 119892 is the gravitational acceleration and 119868119887is the inertia of
the ball In Figure 2 120579119909represents the angle of the load gear
The relationship between 120572 and 120579119909is as follows
sin120572 =2 sin 120579
119909119903arm
119871 tbl (2)
where 119871 tbl is the length of the side of the plate and 119903arm is thelength between the joint and the center of the load gear Therelationship of 120573 and 120579
119910is the same as (2) since both gear
systems have the same hardware and the plate is symmetricalThe numerical values of the constant parameters in theequations of motion and (2) are shown in Table 1 Since 120579
119909
and 120579119910are limited as
minus30 [∘] le 120579
119909 120579119910 le 30 [
∘] (3)
from (2) the working ranges of 120572 and 120573 are
minus53 [∘] le 120572 120573 le 53 [
∘] (4)
If the angular velocities and 120573 are relatively low theapproximations
120573 = 0 2= 0 120573
2= 0 (5)
Journal of Applied Mathematics 3
Table 1 Parameters of the ball and plate system
Parameters Numerical values119898119887
00252 [kg]119903119887
00170 [m]119892 981 [ms2]119868119887
289 times 10minus6 [kgsdotm]
119871 tbl 0275 [m]119903arm 00254 [m]
are often used Linearizing the equations of motion at 120579119909= 0
and 120579119910= 0 we have
(119898119887+119868119887
1199032
119887
) 119887+2119898119887119892119903arm119871 tbl
120579119909= 0
(119898119887+119868119887
1199032
119887
) 119910119887+2119898119887119892119903arm119871 tbl
120579119910= 0
(6)
Since the axes are independent of each other we can focuson one axis for example the119883-axis For the input 120579
119909and the
output 119909119887 the transfer function is given by
119875 (119904) =119883119887(119904)
Θ119909(119904)
=
119870bap
1199042 (7)
where119883119887(119904) = L119909
119887(119905) Θ
119909(119904) = L120579
119909(119905) and
119870bap = minus2119898119887119892119903arm119903
2
119887
119871 tbl (1198981198871199032
119887+ 119868119887) (8)
22 Measurement Noise In this visual feedback system thereis inevitable noise from the camera To examine the noiselevel and frequencies we observed the error signal betweena fixed ball position and a measurement signal The resultsare shown in Figure 3 where the upper part represents a timehistory of the error signal including noise and the lower partrepresents the fast Fourier transform (FFT) analysis of theerror signal The noise level in the error signals is relativelyhigh at frequencies over 20 ([rads])
3 I-PD Control by GKYP Lemma
This section describes an I-PD controller design methodbased on the GKYP lemma The feedback control system isshown in Figure 4 where a filter is introduced into the controlsystem
31 Low-Pass Filter In the previous section we showed thatthemeasurement noise degrades the control performance Toreduce the influence of the noise a low-pass filter is availablein the controller design According to the noise propertiesthat we observed it is sufficient to introduce a first-order low-pass filter into the output of the measurement such that
119865 (119904) =120583
119904 + 120583 (9)
where 120583 ([rads]) is the cut-off frequency
0 1 2 3 4 5 6 7 8Time (s)
Am
plitu
de (m
)
002040608
112141618
2
Frequency (rads)
Pow
er
minus49
minus485
minus48
minus475times10minus3
100 101 102
Figure 3 Time history and FFT analysis results of measumentsignal
r e
minus minus
I-PD controlleru
Actuator Ball and platesystem
Low-passfilter
I
P + D
Figure 4 Feedback control system of the ball and plate
32 I-PD Controller In the standard PID control majorchanges in the reference signals cause large input signals tobe generated by the proportional and derivative actions in thecontroller that saturate the actuator Indeed it is difficult totune the parameters in the PID controller (Figure 4) such thatthe actuator in this visual feedback system is not saturatedThe control system with the I-PD controller (Figure 4) hasa structure whose inner loop includes the proportional andderivative actions [7 8] In this structure the integral actionalone acts on the error signal and prevents large signals beinginput to the actuator The control input 119906 can be written as
119906 = minus119870119901119910 +
119870119894
119904(119903 minus 119910) minus
119870119889119904
120591119904 + 1119910 (10)
where 120591(gt 0) is the parameter to approximate the differen-tiator by a proper transfer function119870
119901119870119894 and119870
119889represent
the proportional integral and derivative gains respectivelyThe open-loop transfer function is
119871 (119904) = 119865 (119904) 119875 (119904)119870 (119904) (11)
where
119870 (119904) = 119870119901+119870119894
119904+
119870119889119904
120591119904 + 1= 119870119901(1 +
1
119879119894119904+
119879119889119904
120591119904 + 1) (12)
119879119894and 119879
119889are the integral time and derivative time respec-
tively It should be noted that the open-loop transfer function
4 Journal of Applied Mathematics
of the I-PD structure is the same as that of the standardPID structure To tune the parameters in the I-PD controllerwe employ an open-loop shaping that realizes a desirablefrequency response of the closed-loop system
33 Generalized KYP Lemma It is known that design spec-ifications for an open-loop transfer function can be reducedto LMIs through the GKYP lemma [9] We briefly review thislemma in the case of continuous-time systems
The design specification consists of a frequency range anda desired property in that range The frequency range can berepresented by
Λ (ΦΨ) = 119904 isin C | 120590 (119904 Φ) = 0 120590 (119904 Ψ) ge 0 (13)
where ΦΨ isin H
120590 (119904 Φ) = [119904lowast1]Φ[
119904
1] = 0 (14)
The equality constraint in (13) distinguishes betweencontinuous-time and discrete-time specifications Since weaddress continuous-time systems in this paper we use Φsuch that
Φ = [0 1
1 0] (15)
The inequality constraint 120590(119904 Ψ) ge 0 in (13) sets a frequencyrangeΩ For example a low frequency range is written as
Ω = 120596 | 120596 le 120603119897 = 120596 | 120590 (119895120596 Ψ) ge 0 (16)
where
Ψ = [0 minus119895
119895 2120603119897
] (17)
Table 2 presents a summary of the choice of Ψ versus a typeof the frequency range Ω where 120603
ℓ 120603ℎ 1206031 and 120603
2are real
positive numbers and 120603119888= (1206031+ 1206032)2 On the other hand
the desired property in a specific frequency range can berepresented by
[119871 (119895120596) 119868]Π [119871lowast(119895120596)
119868] ≺ 0 (18)
where
Π = [Π11
Π12
Π21
Π22
] isin H119898+119901
Π11isin H119901 Π ⪰ 0 (19)
119898 and 119901 are the input and output numbers of 119871(119904) respec-tively For SISO systems consider the requirement that 119871(119895120596)in a frequency range is on the half plane under a straight lineThat is 119871(119895120596) is under the straight line such that
119886R [119871 (119895120596)] + 119887I [119871 (119895120596)] lt 119888 (20)
that is equivalent to (18) with
Π = [0 119886 minus 119895119887
119886 + 119895119887 minus2119888] isin H
2 (21)
Table 2 Relation between Ψ andΩ for continuous time
Ψ Ω
[0 minus119895
119895 2120603ℓ
] 120596 le 120603ℓ(low)
[minus1 119895120603
119888
minus11989512060311988812060311206032
] 1206031le 120596 le 120603
2(middle)
[0 119895
minus119895 minus2120603ℎ
] 120603ℎle 120596 (high)
This requirement is designed to reduce sensitivity in a lowfrequency range Another requirement is that 119871(119895120596) in afrequency range is in the interior of the circle of radius 119903withthe center at 119888 That is 119871(119895120596) is in the circle such that
1003816100381610038161003816119871 (119895120596) minus 1198881003816100381610038161003816
2
lt 1199032 (22)
which is equivalent to (18) with
Π = [1 minus119888
lowast
minus119888 |119888|2minus 1199032] isin H
2 (23)
This requirement is designed to guarantee robustness in ahigh frequency range Under these preparations the gener-alized KYP lemma [9] is expressed as follows
Lemma 1 Let 119871(119904) be 119862(119904119868 minus 119860)minus1119861 + 119863 Λ(ΦΨ) in (13) andΠ in (18) are given Assume that det(119904119868 minus 119860) = 0 for all 119904 isin ΛThen the finite frequency condition
[119871 (119904) 119868]Π [119871lowast(119904)
119868] ≺ 0 forall119904 isin Λ (ΦΨ) (24)
holds if and only if there exist Hermitianmatrices119875 and119876 suchthat the matrix inequality condition
[[[
[
Γ [119861
119863]Π11
Π11[119861
119863]
lowast
minusΠ11
]]]
]
≺ 0 (25)
is satisfied where
Γ = [119860 119868
119862 0] (Φ⊤otimes 119875 + Ψ
⊤otimes 119876) [
119860 119868
119862 0]
⊤
+ [
[
0 119861Π12
Πlowast
12119861⊤119863Π12+ Πlowast
12119863⊤+ Π22
]
]
(26)
Equation (25) is affine with respect to 119861119863 119875119876 andΠ22
In the case where 119861 and119863 have affine design parameters (25)is an LMI
4 Control System Design
41 Filter Design Considering the control performance andthe noise level we set the cut-off frequency in (9) as 120583 =
20 ([rads]) We examine the effectiveness of this filter in
Journal of Applied Mathematics 5
LPF
0 1 2 3 4 5 6 7 8Time (s)
Am
plitu
de (m
)
002040608
112141618
2
Frequency (rads)
Pow
er
minus49
minus485
minus48
minus475
times10minus3
100 101 102
Raw data
Figure 5 Time history and FFT analysis result of measurementsignal with low-pass filter
the same experimental setup as given in Section 22 Figure 5shows the spectral analysis results of the measurement andfiltered signals which are represented by the dotted and solidcurves respectively From these results it can be seen that thenoise at frequencies over 20 ([rads]) has been reduced
42 State-Space Realization of Open-Loop Transfer FunctionTo obtain an LMI based on the GKYP lemma a state-spacerealization of 119871(119904) is required to be affine with respect to aset of the design parameters 120588 = (119870
119901 119870119894 119870119889) If we fix 120591 in
119870(119904) at 10times10minus2 the design parameters appear affinely in thenumerator of119870(119904) Indeed the controllable canonical form of119870(119904) is written as
[119860119896119861119896(120588)
119862119896119863119896(120588)
] =
[[[[[[
[
0 0119870119894
120591
1 minus1
120591119870119894minus119870119889
1205912
0 1 119870119901+119870119889
120591
]]]]]]
]
(27)
Realizations of 119875(119904) and 119865(119904) are also written as
[119860119901
119861119901
119862119901119863119901
] = [
[
0 1 0
0 0 119870bap1 0 0
]
]
(28)
[119860119891
119861119891
119862119891
119863119891
] = [minus120583 1
120583 0] (29)
respectively By combining these realizations (27)ndash(29) weobtain a realization of 119871(119904) as
119871 (119904) = [119860 119861 (120588)
119862 119863 (120588)] (30)
where
119860 = [
[
119860119896
0 0
119862119896119861119901
119860119901
0
119862119896119863119901119861119891119862119901119861119891119860119891
]
]
119861 = [
[
119861119896(120588)
119863119896(120588) 119861119901
119863119896(120588)119863119901119861119891
]
]
119862 = [119862119896119863119901119861119891 119862119901119861119891119862119891] 119863 = 119863
119896(120588)119863119901119863119891
(31)
Consequently the state-space realization of 119871(119904) is affine withrespect to 120588
43 Specifications To shape the Nyquist plot of 119871(119904) werequire the following FDI specifications
minus2R [119871 (119895120596)] +I [119871 (119895120596)] gt 120574ℓforall120596 le 08 (32)
I [119871 (119895120596)] lt 120574119898 25 le
forall120596 le 28 (33)
1003816100381610038161003816119871 (119895120596)1003816100381610038161003816 lt 120574ℎ
forall120596 ge 10 (34)
Specification (32) with a large 120574ℓ(gt 0) ensures sensitivity
reduction in the low frequency range by making the gain of119871(119904) high Specification (33) requires the Nyquist plot to beoutside a circle with its center at the point minus1 + 1198950 so thata certain stability margin is guaranteed Specification (34)with a small 120574
ℎensures robustness against the unmodeled
dynamics that typically exists in the high frequency rangeIn addition to the above basic specifications we also
require the following FDIs that improve the property oftrajectory tracking
4R [119871 (119895120596)] +I [119871 (119895120596)] lt 1205741forall120596 ge 03 (35)
4R [119871 (119895120596)] +I [119871 (119895120596)] gt 1205742forall120596 ge 03 (36)
Since the integral action aloneworks on the error between theoutput and the reference signals the property of trajectorytracking depends directly on the integrator Here we focuson the corner angular frequency 120596
119868by the integral action
in the I-PD controller which is given by 120596119868= 1119879
119894 where
119879119894= 119870119901119870119894 We have a strong integral action and the error is
corrected quickly when the corner angular frequency is highwhile too high a corner angular frequency causes overshootand hunting Thus we impose restrictions for the phase ofthe I-PD controller It should be noted that the phase atlower frequencies is about minus90 ([deg]) while the phase atthe corner angular frequency is about 0 [deg] If we find thefrequency at a specific phase from minus90 ([deg]) to 0 ([deg])the corner angular frequency is greater than that frequencySpecifications (35) and (36) restrict the phase of the I-PDcontroller as well as the open-loop transfer function so thatthe corner angular frequency is greater than the frequency atthe lowest point in the frequency range 03 ([rads])
44 I-PD Controller Design We design an I-PD controller bymaximizing 120574
ℓsubject to Specifications (32)ndash(36) where
(120574119898 120574ℎ 1205741 1205742) = (minus1 05 1 minus100) (37)
6 Journal of Applied Mathematics
0 50 100 150 200
0
50
100
150
Real axis
Imag
inar
y ax
is
The upper limit
Low frequency
The lower limit
minus150
minus100
minus50
minus150minus200 minus100 minus50
Figure 6 Nyquist plot
That is the optimization problem is
max119870119894 119870119901119870119889 120574ℓ
120574ℓ
subject to (19)ndash(36) and (37)
(38)
Each of the design Specifications (32)ndash(36) is reduced to anLMI condition through Lemma 1 with the realization (30)The Specification (32) is modified to 120576 le forall120596 le 08 where 120576 =10 times 10
minus4 because 119871(119904) includes the origin poles that preventus from taking 120596 = 0 Then the LMI optimization problemis to maximize 120574
ℓsubject to all these LMI conditions where
119870119894 119870119901 and 119870
119889are the common decision variables while
1198751 119875
5and 119876
1 119876
5appear in the LMIs as independent
decision variables It should be noted that 120574ℓappears alone in
Π22in the LMI condition (25) corresponding to (36) In this
sense 120574ℓis also an independent decision variable It should
also be noted that 119875 and 119876 in (25) appear in each of the LMIconditions as the independent decision variables
To solve this LMI optimization problem we use YALMIPR20120806 [11] an LMI parser and SPDT3 version 40 [12]an LMI solver on MATLAB R2011b The resulting optimalparameters in the I-PD controller and 120574
ℓare
(119870119901 119870119894 119870119889 120574ℓ) = (92859 107806 4138 119464)
(39)
The Nyquist plots are shown in Figures 6 and 7 where 119871(119895120596)satisfies the design specifications given in Section 43 Since 120574
ℓ
is maximized sensitivity is reduced in the frequency rangeThe Bode plots of 119865(119904)119875(119904) 119871(119904) and 119870(119904) are shown inFigure 8 where the corner angular frequency 120596
119868of the I-PD
controller is larger than 03 ([rads])
5 Simulation Results
The I-PD controller whose design is described in the previoussection was evaluated by a simulation of the step response Tocompare the response with that of a standard PID controller
0 1
0
05
1
15
Real axis
Imag
inar
y ax
is
Middle frequency
High frequency
minus15
minus1
minus05
minus5 minus4 minus3 minus2 minus1
120596 = 102 (rads)
120596 = 30 (rads)
Figure 7 Nyquist plot (magnification)
0
100
200
Gai
n (d
B)
0
90
Phas
e (de
g)
Frequency (rads)
minus300
minus200
minus100
minus270
minus180
minus90
10minus2 100 102 104
P(s) + F(s)
L(s)
K(s)
Frequency (rads)10minus2 100 102 104
Figure 8 Bode plot
we used the PID controller whose gain parameters are thesame as those of the I-PD controller Since both feedbacksystems have the same open-loop transfer function theirfeedback properties must be the same provided that eachinput signal does not saturate The simulation results of thestep response are shown in Figure 9 where the upper andlower parts are the output and input signals respectivelyThe solid curves represent the responses given by the I-PDcontroller while the dashed curves represent those by the PIDcontroller One can see that the input signal given by the PIDcontroller is saturated while that by the I-PD controller is notsaturated and satisfies the limitation (3) The output signalgiven by the I-PD controller settles down to the desired valuewithout any overshoot
It should be noted that the gain parameters in thedesigned controller are not tuned with the I-PD structure
Journal of Applied Mathematics 7
0 1 2 3 4 50
005
01
015
Am
plitu
de (m
)
0 1 2 3 4 5
0102030
Time (s)
Ang
le (d
eg)
I-PDPID
minus30
minus20
minus10
Time (s)
Figure 9 Step response simulation results
0 05 1 15 2 25 3 35 4 45 50
005
01
015
Time (s)
Am
plitu
de (m
)
0 05 1 15 2 25 3 35 4 45 5
0102030
Am
plitu
de (r
ad)
minus30
minus20
minus10
Time (s)
I-PDPID
Figure 10 Step response experimental results
In our experience it is difficult not to saturate the inputlimitation for the PID structure using any design method
6 Experimental Results
This section evaluates the I-PD controller whose designis given in Section 4 through an experiment of the stepresponse The PID controller used in Section 4 was alsoevaluated for comparison The results of trajectory trackingcontrol by the I-PD controller were also evaluated
61 Step Response Experiment The results of the stepresponse experiment are shown in Figure 10 where thedescription of the figure is the same as that of Figure 9 In thisexperiment the influence of the time delay appeared and theinput signals were slightly larger than those obtained in thesimulations The rise and settling time results are howeveralmost the same as those obtained in the simulation
62 Trajectory Tracking Experiments We tested two kindsof trajectories for tracking control a square and a circular
8 Journal of Applied Mathematics
0 002 004
0
002
004
006
Reference
minus006minus006
minus004
minus004
minus002
minus002 0 002 004
0
002
004
006
minus006minus006
minus004
minus004
minus002
minus002
yb
(m)
xb (m)
I-PD controlyb
(m)
xb (m)
ReferenceI-PD control
Figure 11 Experimental results of tracking control
0 5 10 15 20 25 30 35 40 45 50
0005
01
02040
Time (s)
0 5 10 15 20 25 30 35 40 45 50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
minus01minus005
minus40minus20
02040
minus40minus20
0005
01
minus01minus005
Time (s)
Time (s)Time (s)
ReferenceI-PD control
Controller outputActuator output
xb
(m)
120572(d
eg)
yb
(m)
120573(d
eg)
Figure 12 Experimental results of square response
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0005
01
02040
minus01minus005
minus40minus20
02040
minus40minus20
0005
01
minus01minus005
Time (s)Time (s)
Time (s)Time (s)
xb
(m)
120572(d
eg)
yb
(m)
120573(d
eg)
ReferenceI-PD control
Controller outputActuator output
Figure 13 Experimental results of circle response
Journal of Applied Mathematics 9
trajectory The side of the square trajectory was 01 ([m])and the radius of the circular trajectory was 005 ([m]) Theresults are shown in Figure 11 where the left part shows atrajectory of the square trajectory tracking control and theright part shows a trajectory of the circular trajectory trackingcontrol The time histories of the ball and input angles areshown in Figures 12 and 13 In the square trajectory trackingcontrol experiment the responses were similar to those inthe step response experiment except for a slight vibrationSuch vibration phenomena are noticeable in the responsesof circular trajectory tracking control in particular the casewhen the input signal is relatively small The reason for thesephenomena could be the friction of the ball against the plateor a backlash of the gear system
7 Conclusions
This paper applied the GKYP lemma to an open-looptransfer function including an I-PD controller and a noisereduction filter for the ball and plate systemThemultiple FDIspecifications for the finite frequency ranges were satisfiedby a solution of the LMI optimization problem The solutionincludes the optimal parameters in the I-PD controller Thefirst-order low-pass filter reduced the noise in the high fre-quency range and improved the steady-state response Bothsimulations and experiments evaluated the effectiveness ofthe designmethod by comparing the standard PID controller
The PI-D (proportional integral-derivative) control sys-tem which moved the derivative controller to the innerfeedback loop also has the same open-loop transfer functionas the standard PID controller Thus the approach in thispaper can also be applied to the PI-D controller
References
[1] MMoarref M Saadat and G Vossoughi ldquoMechatronic designand position control of a novel ball and plate systemrdquo inProceedings of the Mediterranean Conference on Control andAutomation (MEDrsquo 08) pp 1071ndash1076 June 2008
[2] P Kokotovic ldquoThe joy of feedback nonlinear and adaptiverdquoIEEE Control Systems Magazine vol 12 no 3 pp 7ndash17 1992
[3] F Zheng X Li X Qian and S Wang ldquoModeling and PIDneural network research for the ball and plate systemrdquo inProceedings of the International Conference on Electronics Com-munications and Control (ICECC rsquo11) pp 331ndash334 September2011
[4] X Fan N Zhang and S Teng ldquoTrajectory planning andtracking of ball andplate systemusing hierarchical fuzzy controlschemerdquo Fuzzy Sets and Systems vol 144 no 2 pp 297ndash3122004
[5] M Bai Y Tian and Y Wang ldquoDecoupled fuzzy sliding modecontrol to ball and plate systemrdquo in Proceedings of the 2ndInternational Conference on Intelligent Control and InformationProcessing (ICICIP rsquo11) pp 685ndash690 July 2011
[6] F Borrelli Constrained Optimal Control of Linear and HybridSystems vol 290 of Lecture Notes in Control and InformationSciences Springer 2003
[7] K J Astrom and T Hagglund Advanced PID Control Instru-mentation Systems 2005
[8] Q Li and Z Kemin Introduction to Feedback Control PrenticeHall 2009
[9] T Iwasaki and S Hara ldquoGeneralized KYP lemma unifiedfrequency domain inequalities with design applicationsrdquo IEEETransactions on Automatic Control vol 50 no 1 pp 41ndash592005
[10] A Rantzer ldquoOn the Kalman-Yakubovich-Popov lemmardquo Sys-tems amp Control Letters vol 28 no 1 pp 7ndash10 1996
[11] J Lofberg ldquoYALMIP a toolbox for modeling and optimizationin MATLABrdquo in Proceedings of the IEEE International Sympo-sium on Computer-Aided Control System Design (CACSD rsquo11)Taipei Taiwan 2004
[12] K C TohM J Todd and RH Tutuncu ldquoSDPT3mdashaMATLABsoftware package for semidefinite programming version 13rdquoOptimization Methods and Software vol 11 no 1 pp 545ndash5811999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 3
Table 1 Parameters of the ball and plate system
Parameters Numerical values119898119887
00252 [kg]119903119887
00170 [m]119892 981 [ms2]119868119887
289 times 10minus6 [kgsdotm]
119871 tbl 0275 [m]119903arm 00254 [m]
are often used Linearizing the equations of motion at 120579119909= 0
and 120579119910= 0 we have
(119898119887+119868119887
1199032
119887
) 119887+2119898119887119892119903arm119871 tbl
120579119909= 0
(119898119887+119868119887
1199032
119887
) 119910119887+2119898119887119892119903arm119871 tbl
120579119910= 0
(6)
Since the axes are independent of each other we can focuson one axis for example the119883-axis For the input 120579
119909and the
output 119909119887 the transfer function is given by
119875 (119904) =119883119887(119904)
Θ119909(119904)
=
119870bap
1199042 (7)
where119883119887(119904) = L119909
119887(119905) Θ
119909(119904) = L120579
119909(119905) and
119870bap = minus2119898119887119892119903arm119903
2
119887
119871 tbl (1198981198871199032
119887+ 119868119887) (8)
22 Measurement Noise In this visual feedback system thereis inevitable noise from the camera To examine the noiselevel and frequencies we observed the error signal betweena fixed ball position and a measurement signal The resultsare shown in Figure 3 where the upper part represents a timehistory of the error signal including noise and the lower partrepresents the fast Fourier transform (FFT) analysis of theerror signal The noise level in the error signals is relativelyhigh at frequencies over 20 ([rads])
3 I-PD Control by GKYP Lemma
This section describes an I-PD controller design methodbased on the GKYP lemma The feedback control system isshown in Figure 4 where a filter is introduced into the controlsystem
31 Low-Pass Filter In the previous section we showed thatthemeasurement noise degrades the control performance Toreduce the influence of the noise a low-pass filter is availablein the controller design According to the noise propertiesthat we observed it is sufficient to introduce a first-order low-pass filter into the output of the measurement such that
119865 (119904) =120583
119904 + 120583 (9)
where 120583 ([rads]) is the cut-off frequency
0 1 2 3 4 5 6 7 8Time (s)
Am
plitu
de (m
)
002040608
112141618
2
Frequency (rads)
Pow
er
minus49
minus485
minus48
minus475times10minus3
100 101 102
Figure 3 Time history and FFT analysis results of measumentsignal
r e
minus minus
I-PD controlleru
Actuator Ball and platesystem
Low-passfilter
I
P + D
Figure 4 Feedback control system of the ball and plate
32 I-PD Controller In the standard PID control majorchanges in the reference signals cause large input signals tobe generated by the proportional and derivative actions in thecontroller that saturate the actuator Indeed it is difficult totune the parameters in the PID controller (Figure 4) such thatthe actuator in this visual feedback system is not saturatedThe control system with the I-PD controller (Figure 4) hasa structure whose inner loop includes the proportional andderivative actions [7 8] In this structure the integral actionalone acts on the error signal and prevents large signals beinginput to the actuator The control input 119906 can be written as
119906 = minus119870119901119910 +
119870119894
119904(119903 minus 119910) minus
119870119889119904
120591119904 + 1119910 (10)
where 120591(gt 0) is the parameter to approximate the differen-tiator by a proper transfer function119870
119901119870119894 and119870
119889represent
the proportional integral and derivative gains respectivelyThe open-loop transfer function is
119871 (119904) = 119865 (119904) 119875 (119904)119870 (119904) (11)
where
119870 (119904) = 119870119901+119870119894
119904+
119870119889119904
120591119904 + 1= 119870119901(1 +
1
119879119894119904+
119879119889119904
120591119904 + 1) (12)
119879119894and 119879
119889are the integral time and derivative time respec-
tively It should be noted that the open-loop transfer function
4 Journal of Applied Mathematics
of the I-PD structure is the same as that of the standardPID structure To tune the parameters in the I-PD controllerwe employ an open-loop shaping that realizes a desirablefrequency response of the closed-loop system
33 Generalized KYP Lemma It is known that design spec-ifications for an open-loop transfer function can be reducedto LMIs through the GKYP lemma [9] We briefly review thislemma in the case of continuous-time systems
The design specification consists of a frequency range anda desired property in that range The frequency range can berepresented by
Λ (ΦΨ) = 119904 isin C | 120590 (119904 Φ) = 0 120590 (119904 Ψ) ge 0 (13)
where ΦΨ isin H
120590 (119904 Φ) = [119904lowast1]Φ[
119904
1] = 0 (14)
The equality constraint in (13) distinguishes betweencontinuous-time and discrete-time specifications Since weaddress continuous-time systems in this paper we use Φsuch that
Φ = [0 1
1 0] (15)
The inequality constraint 120590(119904 Ψ) ge 0 in (13) sets a frequencyrangeΩ For example a low frequency range is written as
Ω = 120596 | 120596 le 120603119897 = 120596 | 120590 (119895120596 Ψ) ge 0 (16)
where
Ψ = [0 minus119895
119895 2120603119897
] (17)
Table 2 presents a summary of the choice of Ψ versus a typeof the frequency range Ω where 120603
ℓ 120603ℎ 1206031 and 120603
2are real
positive numbers and 120603119888= (1206031+ 1206032)2 On the other hand
the desired property in a specific frequency range can berepresented by
[119871 (119895120596) 119868]Π [119871lowast(119895120596)
119868] ≺ 0 (18)
where
Π = [Π11
Π12
Π21
Π22
] isin H119898+119901
Π11isin H119901 Π ⪰ 0 (19)
119898 and 119901 are the input and output numbers of 119871(119904) respec-tively For SISO systems consider the requirement that 119871(119895120596)in a frequency range is on the half plane under a straight lineThat is 119871(119895120596) is under the straight line such that
119886R [119871 (119895120596)] + 119887I [119871 (119895120596)] lt 119888 (20)
that is equivalent to (18) with
Π = [0 119886 minus 119895119887
119886 + 119895119887 minus2119888] isin H
2 (21)
Table 2 Relation between Ψ andΩ for continuous time
Ψ Ω
[0 minus119895
119895 2120603ℓ
] 120596 le 120603ℓ(low)
[minus1 119895120603
119888
minus11989512060311988812060311206032
] 1206031le 120596 le 120603
2(middle)
[0 119895
minus119895 minus2120603ℎ
] 120603ℎle 120596 (high)
This requirement is designed to reduce sensitivity in a lowfrequency range Another requirement is that 119871(119895120596) in afrequency range is in the interior of the circle of radius 119903withthe center at 119888 That is 119871(119895120596) is in the circle such that
1003816100381610038161003816119871 (119895120596) minus 1198881003816100381610038161003816
2
lt 1199032 (22)
which is equivalent to (18) with
Π = [1 minus119888
lowast
minus119888 |119888|2minus 1199032] isin H
2 (23)
This requirement is designed to guarantee robustness in ahigh frequency range Under these preparations the gener-alized KYP lemma [9] is expressed as follows
Lemma 1 Let 119871(119904) be 119862(119904119868 minus 119860)minus1119861 + 119863 Λ(ΦΨ) in (13) andΠ in (18) are given Assume that det(119904119868 minus 119860) = 0 for all 119904 isin ΛThen the finite frequency condition
[119871 (119904) 119868]Π [119871lowast(119904)
119868] ≺ 0 forall119904 isin Λ (ΦΨ) (24)
holds if and only if there exist Hermitianmatrices119875 and119876 suchthat the matrix inequality condition
[[[
[
Γ [119861
119863]Π11
Π11[119861
119863]
lowast
minusΠ11
]]]
]
≺ 0 (25)
is satisfied where
Γ = [119860 119868
119862 0] (Φ⊤otimes 119875 + Ψ
⊤otimes 119876) [
119860 119868
119862 0]
⊤
+ [
[
0 119861Π12
Πlowast
12119861⊤119863Π12+ Πlowast
12119863⊤+ Π22
]
]
(26)
Equation (25) is affine with respect to 119861119863 119875119876 andΠ22
In the case where 119861 and119863 have affine design parameters (25)is an LMI
4 Control System Design
41 Filter Design Considering the control performance andthe noise level we set the cut-off frequency in (9) as 120583 =
20 ([rads]) We examine the effectiveness of this filter in
Journal of Applied Mathematics 5
LPF
0 1 2 3 4 5 6 7 8Time (s)
Am
plitu
de (m
)
002040608
112141618
2
Frequency (rads)
Pow
er
minus49
minus485
minus48
minus475
times10minus3
100 101 102
Raw data
Figure 5 Time history and FFT analysis result of measurementsignal with low-pass filter
the same experimental setup as given in Section 22 Figure 5shows the spectral analysis results of the measurement andfiltered signals which are represented by the dotted and solidcurves respectively From these results it can be seen that thenoise at frequencies over 20 ([rads]) has been reduced
42 State-Space Realization of Open-Loop Transfer FunctionTo obtain an LMI based on the GKYP lemma a state-spacerealization of 119871(119904) is required to be affine with respect to aset of the design parameters 120588 = (119870
119901 119870119894 119870119889) If we fix 120591 in
119870(119904) at 10times10minus2 the design parameters appear affinely in thenumerator of119870(119904) Indeed the controllable canonical form of119870(119904) is written as
[119860119896119861119896(120588)
119862119896119863119896(120588)
] =
[[[[[[
[
0 0119870119894
120591
1 minus1
120591119870119894minus119870119889
1205912
0 1 119870119901+119870119889
120591
]]]]]]
]
(27)
Realizations of 119875(119904) and 119865(119904) are also written as
[119860119901
119861119901
119862119901119863119901
] = [
[
0 1 0
0 0 119870bap1 0 0
]
]
(28)
[119860119891
119861119891
119862119891
119863119891
] = [minus120583 1
120583 0] (29)
respectively By combining these realizations (27)ndash(29) weobtain a realization of 119871(119904) as
119871 (119904) = [119860 119861 (120588)
119862 119863 (120588)] (30)
where
119860 = [
[
119860119896
0 0
119862119896119861119901
119860119901
0
119862119896119863119901119861119891119862119901119861119891119860119891
]
]
119861 = [
[
119861119896(120588)
119863119896(120588) 119861119901
119863119896(120588)119863119901119861119891
]
]
119862 = [119862119896119863119901119861119891 119862119901119861119891119862119891] 119863 = 119863
119896(120588)119863119901119863119891
(31)
Consequently the state-space realization of 119871(119904) is affine withrespect to 120588
43 Specifications To shape the Nyquist plot of 119871(119904) werequire the following FDI specifications
minus2R [119871 (119895120596)] +I [119871 (119895120596)] gt 120574ℓforall120596 le 08 (32)
I [119871 (119895120596)] lt 120574119898 25 le
forall120596 le 28 (33)
1003816100381610038161003816119871 (119895120596)1003816100381610038161003816 lt 120574ℎ
forall120596 ge 10 (34)
Specification (32) with a large 120574ℓ(gt 0) ensures sensitivity
reduction in the low frequency range by making the gain of119871(119904) high Specification (33) requires the Nyquist plot to beoutside a circle with its center at the point minus1 + 1198950 so thata certain stability margin is guaranteed Specification (34)with a small 120574
ℎensures robustness against the unmodeled
dynamics that typically exists in the high frequency rangeIn addition to the above basic specifications we also
require the following FDIs that improve the property oftrajectory tracking
4R [119871 (119895120596)] +I [119871 (119895120596)] lt 1205741forall120596 ge 03 (35)
4R [119871 (119895120596)] +I [119871 (119895120596)] gt 1205742forall120596 ge 03 (36)
Since the integral action aloneworks on the error between theoutput and the reference signals the property of trajectorytracking depends directly on the integrator Here we focuson the corner angular frequency 120596
119868by the integral action
in the I-PD controller which is given by 120596119868= 1119879
119894 where
119879119894= 119870119901119870119894 We have a strong integral action and the error is
corrected quickly when the corner angular frequency is highwhile too high a corner angular frequency causes overshootand hunting Thus we impose restrictions for the phase ofthe I-PD controller It should be noted that the phase atlower frequencies is about minus90 ([deg]) while the phase atthe corner angular frequency is about 0 [deg] If we find thefrequency at a specific phase from minus90 ([deg]) to 0 ([deg])the corner angular frequency is greater than that frequencySpecifications (35) and (36) restrict the phase of the I-PDcontroller as well as the open-loop transfer function so thatthe corner angular frequency is greater than the frequency atthe lowest point in the frequency range 03 ([rads])
44 I-PD Controller Design We design an I-PD controller bymaximizing 120574
ℓsubject to Specifications (32)ndash(36) where
(120574119898 120574ℎ 1205741 1205742) = (minus1 05 1 minus100) (37)
6 Journal of Applied Mathematics
0 50 100 150 200
0
50
100
150
Real axis
Imag
inar
y ax
is
The upper limit
Low frequency
The lower limit
minus150
minus100
minus50
minus150minus200 minus100 minus50
Figure 6 Nyquist plot
That is the optimization problem is
max119870119894 119870119901119870119889 120574ℓ
120574ℓ
subject to (19)ndash(36) and (37)
(38)
Each of the design Specifications (32)ndash(36) is reduced to anLMI condition through Lemma 1 with the realization (30)The Specification (32) is modified to 120576 le forall120596 le 08 where 120576 =10 times 10
minus4 because 119871(119904) includes the origin poles that preventus from taking 120596 = 0 Then the LMI optimization problemis to maximize 120574
ℓsubject to all these LMI conditions where
119870119894 119870119901 and 119870
119889are the common decision variables while
1198751 119875
5and 119876
1 119876
5appear in the LMIs as independent
decision variables It should be noted that 120574ℓappears alone in
Π22in the LMI condition (25) corresponding to (36) In this
sense 120574ℓis also an independent decision variable It should
also be noted that 119875 and 119876 in (25) appear in each of the LMIconditions as the independent decision variables
To solve this LMI optimization problem we use YALMIPR20120806 [11] an LMI parser and SPDT3 version 40 [12]an LMI solver on MATLAB R2011b The resulting optimalparameters in the I-PD controller and 120574
ℓare
(119870119901 119870119894 119870119889 120574ℓ) = (92859 107806 4138 119464)
(39)
The Nyquist plots are shown in Figures 6 and 7 where 119871(119895120596)satisfies the design specifications given in Section 43 Since 120574
ℓ
is maximized sensitivity is reduced in the frequency rangeThe Bode plots of 119865(119904)119875(119904) 119871(119904) and 119870(119904) are shown inFigure 8 where the corner angular frequency 120596
119868of the I-PD
controller is larger than 03 ([rads])
5 Simulation Results
The I-PD controller whose design is described in the previoussection was evaluated by a simulation of the step response Tocompare the response with that of a standard PID controller
0 1
0
05
1
15
Real axis
Imag
inar
y ax
is
Middle frequency
High frequency
minus15
minus1
minus05
minus5 minus4 minus3 minus2 minus1
120596 = 102 (rads)
120596 = 30 (rads)
Figure 7 Nyquist plot (magnification)
0
100
200
Gai
n (d
B)
0
90
Phas
e (de
g)
Frequency (rads)
minus300
minus200
minus100
minus270
minus180
minus90
10minus2 100 102 104
P(s) + F(s)
L(s)
K(s)
Frequency (rads)10minus2 100 102 104
Figure 8 Bode plot
we used the PID controller whose gain parameters are thesame as those of the I-PD controller Since both feedbacksystems have the same open-loop transfer function theirfeedback properties must be the same provided that eachinput signal does not saturate The simulation results of thestep response are shown in Figure 9 where the upper andlower parts are the output and input signals respectivelyThe solid curves represent the responses given by the I-PDcontroller while the dashed curves represent those by the PIDcontroller One can see that the input signal given by the PIDcontroller is saturated while that by the I-PD controller is notsaturated and satisfies the limitation (3) The output signalgiven by the I-PD controller settles down to the desired valuewithout any overshoot
It should be noted that the gain parameters in thedesigned controller are not tuned with the I-PD structure
Journal of Applied Mathematics 7
0 1 2 3 4 50
005
01
015
Am
plitu
de (m
)
0 1 2 3 4 5
0102030
Time (s)
Ang
le (d
eg)
I-PDPID
minus30
minus20
minus10
Time (s)
Figure 9 Step response simulation results
0 05 1 15 2 25 3 35 4 45 50
005
01
015
Time (s)
Am
plitu
de (m
)
0 05 1 15 2 25 3 35 4 45 5
0102030
Am
plitu
de (r
ad)
minus30
minus20
minus10
Time (s)
I-PDPID
Figure 10 Step response experimental results
In our experience it is difficult not to saturate the inputlimitation for the PID structure using any design method
6 Experimental Results
This section evaluates the I-PD controller whose designis given in Section 4 through an experiment of the stepresponse The PID controller used in Section 4 was alsoevaluated for comparison The results of trajectory trackingcontrol by the I-PD controller were also evaluated
61 Step Response Experiment The results of the stepresponse experiment are shown in Figure 10 where thedescription of the figure is the same as that of Figure 9 In thisexperiment the influence of the time delay appeared and theinput signals were slightly larger than those obtained in thesimulations The rise and settling time results are howeveralmost the same as those obtained in the simulation
62 Trajectory Tracking Experiments We tested two kindsof trajectories for tracking control a square and a circular
8 Journal of Applied Mathematics
0 002 004
0
002
004
006
Reference
minus006minus006
minus004
minus004
minus002
minus002 0 002 004
0
002
004
006
minus006minus006
minus004
minus004
minus002
minus002
yb
(m)
xb (m)
I-PD controlyb
(m)
xb (m)
ReferenceI-PD control
Figure 11 Experimental results of tracking control
0 5 10 15 20 25 30 35 40 45 50
0005
01
02040
Time (s)
0 5 10 15 20 25 30 35 40 45 50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
minus01minus005
minus40minus20
02040
minus40minus20
0005
01
minus01minus005
Time (s)
Time (s)Time (s)
ReferenceI-PD control
Controller outputActuator output
xb
(m)
120572(d
eg)
yb
(m)
120573(d
eg)
Figure 12 Experimental results of square response
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0005
01
02040
minus01minus005
minus40minus20
02040
minus40minus20
0005
01
minus01minus005
Time (s)Time (s)
Time (s)Time (s)
xb
(m)
120572(d
eg)
yb
(m)
120573(d
eg)
ReferenceI-PD control
Controller outputActuator output
Figure 13 Experimental results of circle response
Journal of Applied Mathematics 9
trajectory The side of the square trajectory was 01 ([m])and the radius of the circular trajectory was 005 ([m]) Theresults are shown in Figure 11 where the left part shows atrajectory of the square trajectory tracking control and theright part shows a trajectory of the circular trajectory trackingcontrol The time histories of the ball and input angles areshown in Figures 12 and 13 In the square trajectory trackingcontrol experiment the responses were similar to those inthe step response experiment except for a slight vibrationSuch vibration phenomena are noticeable in the responsesof circular trajectory tracking control in particular the casewhen the input signal is relatively small The reason for thesephenomena could be the friction of the ball against the plateor a backlash of the gear system
7 Conclusions
This paper applied the GKYP lemma to an open-looptransfer function including an I-PD controller and a noisereduction filter for the ball and plate systemThemultiple FDIspecifications for the finite frequency ranges were satisfiedby a solution of the LMI optimization problem The solutionincludes the optimal parameters in the I-PD controller Thefirst-order low-pass filter reduced the noise in the high fre-quency range and improved the steady-state response Bothsimulations and experiments evaluated the effectiveness ofthe designmethod by comparing the standard PID controller
The PI-D (proportional integral-derivative) control sys-tem which moved the derivative controller to the innerfeedback loop also has the same open-loop transfer functionas the standard PID controller Thus the approach in thispaper can also be applied to the PI-D controller
References
[1] MMoarref M Saadat and G Vossoughi ldquoMechatronic designand position control of a novel ball and plate systemrdquo inProceedings of the Mediterranean Conference on Control andAutomation (MEDrsquo 08) pp 1071ndash1076 June 2008
[2] P Kokotovic ldquoThe joy of feedback nonlinear and adaptiverdquoIEEE Control Systems Magazine vol 12 no 3 pp 7ndash17 1992
[3] F Zheng X Li X Qian and S Wang ldquoModeling and PIDneural network research for the ball and plate systemrdquo inProceedings of the International Conference on Electronics Com-munications and Control (ICECC rsquo11) pp 331ndash334 September2011
[4] X Fan N Zhang and S Teng ldquoTrajectory planning andtracking of ball andplate systemusing hierarchical fuzzy controlschemerdquo Fuzzy Sets and Systems vol 144 no 2 pp 297ndash3122004
[5] M Bai Y Tian and Y Wang ldquoDecoupled fuzzy sliding modecontrol to ball and plate systemrdquo in Proceedings of the 2ndInternational Conference on Intelligent Control and InformationProcessing (ICICIP rsquo11) pp 685ndash690 July 2011
[6] F Borrelli Constrained Optimal Control of Linear and HybridSystems vol 290 of Lecture Notes in Control and InformationSciences Springer 2003
[7] K J Astrom and T Hagglund Advanced PID Control Instru-mentation Systems 2005
[8] Q Li and Z Kemin Introduction to Feedback Control PrenticeHall 2009
[9] T Iwasaki and S Hara ldquoGeneralized KYP lemma unifiedfrequency domain inequalities with design applicationsrdquo IEEETransactions on Automatic Control vol 50 no 1 pp 41ndash592005
[10] A Rantzer ldquoOn the Kalman-Yakubovich-Popov lemmardquo Sys-tems amp Control Letters vol 28 no 1 pp 7ndash10 1996
[11] J Lofberg ldquoYALMIP a toolbox for modeling and optimizationin MATLABrdquo in Proceedings of the IEEE International Sympo-sium on Computer-Aided Control System Design (CACSD rsquo11)Taipei Taiwan 2004
[12] K C TohM J Todd and RH Tutuncu ldquoSDPT3mdashaMATLABsoftware package for semidefinite programming version 13rdquoOptimization Methods and Software vol 11 no 1 pp 545ndash5811999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
4 Journal of Applied Mathematics
of the I-PD structure is the same as that of the standardPID structure To tune the parameters in the I-PD controllerwe employ an open-loop shaping that realizes a desirablefrequency response of the closed-loop system
33 Generalized KYP Lemma It is known that design spec-ifications for an open-loop transfer function can be reducedto LMIs through the GKYP lemma [9] We briefly review thislemma in the case of continuous-time systems
The design specification consists of a frequency range anda desired property in that range The frequency range can berepresented by
Λ (ΦΨ) = 119904 isin C | 120590 (119904 Φ) = 0 120590 (119904 Ψ) ge 0 (13)
where ΦΨ isin H
120590 (119904 Φ) = [119904lowast1]Φ[
119904
1] = 0 (14)
The equality constraint in (13) distinguishes betweencontinuous-time and discrete-time specifications Since weaddress continuous-time systems in this paper we use Φsuch that
Φ = [0 1
1 0] (15)
The inequality constraint 120590(119904 Ψ) ge 0 in (13) sets a frequencyrangeΩ For example a low frequency range is written as
Ω = 120596 | 120596 le 120603119897 = 120596 | 120590 (119895120596 Ψ) ge 0 (16)
where
Ψ = [0 minus119895
119895 2120603119897
] (17)
Table 2 presents a summary of the choice of Ψ versus a typeof the frequency range Ω where 120603
ℓ 120603ℎ 1206031 and 120603
2are real
positive numbers and 120603119888= (1206031+ 1206032)2 On the other hand
the desired property in a specific frequency range can berepresented by
[119871 (119895120596) 119868]Π [119871lowast(119895120596)
119868] ≺ 0 (18)
where
Π = [Π11
Π12
Π21
Π22
] isin H119898+119901
Π11isin H119901 Π ⪰ 0 (19)
119898 and 119901 are the input and output numbers of 119871(119904) respec-tively For SISO systems consider the requirement that 119871(119895120596)in a frequency range is on the half plane under a straight lineThat is 119871(119895120596) is under the straight line such that
119886R [119871 (119895120596)] + 119887I [119871 (119895120596)] lt 119888 (20)
that is equivalent to (18) with
Π = [0 119886 minus 119895119887
119886 + 119895119887 minus2119888] isin H
2 (21)
Table 2 Relation between Ψ andΩ for continuous time
Ψ Ω
[0 minus119895
119895 2120603ℓ
] 120596 le 120603ℓ(low)
[minus1 119895120603
119888
minus11989512060311988812060311206032
] 1206031le 120596 le 120603
2(middle)
[0 119895
minus119895 minus2120603ℎ
] 120603ℎle 120596 (high)
This requirement is designed to reduce sensitivity in a lowfrequency range Another requirement is that 119871(119895120596) in afrequency range is in the interior of the circle of radius 119903withthe center at 119888 That is 119871(119895120596) is in the circle such that
1003816100381610038161003816119871 (119895120596) minus 1198881003816100381610038161003816
2
lt 1199032 (22)
which is equivalent to (18) with
Π = [1 minus119888
lowast
minus119888 |119888|2minus 1199032] isin H
2 (23)
This requirement is designed to guarantee robustness in ahigh frequency range Under these preparations the gener-alized KYP lemma [9] is expressed as follows
Lemma 1 Let 119871(119904) be 119862(119904119868 minus 119860)minus1119861 + 119863 Λ(ΦΨ) in (13) andΠ in (18) are given Assume that det(119904119868 minus 119860) = 0 for all 119904 isin ΛThen the finite frequency condition
[119871 (119904) 119868]Π [119871lowast(119904)
119868] ≺ 0 forall119904 isin Λ (ΦΨ) (24)
holds if and only if there exist Hermitianmatrices119875 and119876 suchthat the matrix inequality condition
[[[
[
Γ [119861
119863]Π11
Π11[119861
119863]
lowast
minusΠ11
]]]
]
≺ 0 (25)
is satisfied where
Γ = [119860 119868
119862 0] (Φ⊤otimes 119875 + Ψ
⊤otimes 119876) [
119860 119868
119862 0]
⊤
+ [
[
0 119861Π12
Πlowast
12119861⊤119863Π12+ Πlowast
12119863⊤+ Π22
]
]
(26)
Equation (25) is affine with respect to 119861119863 119875119876 andΠ22
In the case where 119861 and119863 have affine design parameters (25)is an LMI
4 Control System Design
41 Filter Design Considering the control performance andthe noise level we set the cut-off frequency in (9) as 120583 =
20 ([rads]) We examine the effectiveness of this filter in
Journal of Applied Mathematics 5
LPF
0 1 2 3 4 5 6 7 8Time (s)
Am
plitu
de (m
)
002040608
112141618
2
Frequency (rads)
Pow
er
minus49
minus485
minus48
minus475
times10minus3
100 101 102
Raw data
Figure 5 Time history and FFT analysis result of measurementsignal with low-pass filter
the same experimental setup as given in Section 22 Figure 5shows the spectral analysis results of the measurement andfiltered signals which are represented by the dotted and solidcurves respectively From these results it can be seen that thenoise at frequencies over 20 ([rads]) has been reduced
42 State-Space Realization of Open-Loop Transfer FunctionTo obtain an LMI based on the GKYP lemma a state-spacerealization of 119871(119904) is required to be affine with respect to aset of the design parameters 120588 = (119870
119901 119870119894 119870119889) If we fix 120591 in
119870(119904) at 10times10minus2 the design parameters appear affinely in thenumerator of119870(119904) Indeed the controllable canonical form of119870(119904) is written as
[119860119896119861119896(120588)
119862119896119863119896(120588)
] =
[[[[[[
[
0 0119870119894
120591
1 minus1
120591119870119894minus119870119889
1205912
0 1 119870119901+119870119889
120591
]]]]]]
]
(27)
Realizations of 119875(119904) and 119865(119904) are also written as
[119860119901
119861119901
119862119901119863119901
] = [
[
0 1 0
0 0 119870bap1 0 0
]
]
(28)
[119860119891
119861119891
119862119891
119863119891
] = [minus120583 1
120583 0] (29)
respectively By combining these realizations (27)ndash(29) weobtain a realization of 119871(119904) as
119871 (119904) = [119860 119861 (120588)
119862 119863 (120588)] (30)
where
119860 = [
[
119860119896
0 0
119862119896119861119901
119860119901
0
119862119896119863119901119861119891119862119901119861119891119860119891
]
]
119861 = [
[
119861119896(120588)
119863119896(120588) 119861119901
119863119896(120588)119863119901119861119891
]
]
119862 = [119862119896119863119901119861119891 119862119901119861119891119862119891] 119863 = 119863
119896(120588)119863119901119863119891
(31)
Consequently the state-space realization of 119871(119904) is affine withrespect to 120588
43 Specifications To shape the Nyquist plot of 119871(119904) werequire the following FDI specifications
minus2R [119871 (119895120596)] +I [119871 (119895120596)] gt 120574ℓforall120596 le 08 (32)
I [119871 (119895120596)] lt 120574119898 25 le
forall120596 le 28 (33)
1003816100381610038161003816119871 (119895120596)1003816100381610038161003816 lt 120574ℎ
forall120596 ge 10 (34)
Specification (32) with a large 120574ℓ(gt 0) ensures sensitivity
reduction in the low frequency range by making the gain of119871(119904) high Specification (33) requires the Nyquist plot to beoutside a circle with its center at the point minus1 + 1198950 so thata certain stability margin is guaranteed Specification (34)with a small 120574
ℎensures robustness against the unmodeled
dynamics that typically exists in the high frequency rangeIn addition to the above basic specifications we also
require the following FDIs that improve the property oftrajectory tracking
4R [119871 (119895120596)] +I [119871 (119895120596)] lt 1205741forall120596 ge 03 (35)
4R [119871 (119895120596)] +I [119871 (119895120596)] gt 1205742forall120596 ge 03 (36)
Since the integral action aloneworks on the error between theoutput and the reference signals the property of trajectorytracking depends directly on the integrator Here we focuson the corner angular frequency 120596
119868by the integral action
in the I-PD controller which is given by 120596119868= 1119879
119894 where
119879119894= 119870119901119870119894 We have a strong integral action and the error is
corrected quickly when the corner angular frequency is highwhile too high a corner angular frequency causes overshootand hunting Thus we impose restrictions for the phase ofthe I-PD controller It should be noted that the phase atlower frequencies is about minus90 ([deg]) while the phase atthe corner angular frequency is about 0 [deg] If we find thefrequency at a specific phase from minus90 ([deg]) to 0 ([deg])the corner angular frequency is greater than that frequencySpecifications (35) and (36) restrict the phase of the I-PDcontroller as well as the open-loop transfer function so thatthe corner angular frequency is greater than the frequency atthe lowest point in the frequency range 03 ([rads])
44 I-PD Controller Design We design an I-PD controller bymaximizing 120574
ℓsubject to Specifications (32)ndash(36) where
(120574119898 120574ℎ 1205741 1205742) = (minus1 05 1 minus100) (37)
6 Journal of Applied Mathematics
0 50 100 150 200
0
50
100
150
Real axis
Imag
inar
y ax
is
The upper limit
Low frequency
The lower limit
minus150
minus100
minus50
minus150minus200 minus100 minus50
Figure 6 Nyquist plot
That is the optimization problem is
max119870119894 119870119901119870119889 120574ℓ
120574ℓ
subject to (19)ndash(36) and (37)
(38)
Each of the design Specifications (32)ndash(36) is reduced to anLMI condition through Lemma 1 with the realization (30)The Specification (32) is modified to 120576 le forall120596 le 08 where 120576 =10 times 10
minus4 because 119871(119904) includes the origin poles that preventus from taking 120596 = 0 Then the LMI optimization problemis to maximize 120574
ℓsubject to all these LMI conditions where
119870119894 119870119901 and 119870
119889are the common decision variables while
1198751 119875
5and 119876
1 119876
5appear in the LMIs as independent
decision variables It should be noted that 120574ℓappears alone in
Π22in the LMI condition (25) corresponding to (36) In this
sense 120574ℓis also an independent decision variable It should
also be noted that 119875 and 119876 in (25) appear in each of the LMIconditions as the independent decision variables
To solve this LMI optimization problem we use YALMIPR20120806 [11] an LMI parser and SPDT3 version 40 [12]an LMI solver on MATLAB R2011b The resulting optimalparameters in the I-PD controller and 120574
ℓare
(119870119901 119870119894 119870119889 120574ℓ) = (92859 107806 4138 119464)
(39)
The Nyquist plots are shown in Figures 6 and 7 where 119871(119895120596)satisfies the design specifications given in Section 43 Since 120574
ℓ
is maximized sensitivity is reduced in the frequency rangeThe Bode plots of 119865(119904)119875(119904) 119871(119904) and 119870(119904) are shown inFigure 8 where the corner angular frequency 120596
119868of the I-PD
controller is larger than 03 ([rads])
5 Simulation Results
The I-PD controller whose design is described in the previoussection was evaluated by a simulation of the step response Tocompare the response with that of a standard PID controller
0 1
0
05
1
15
Real axis
Imag
inar
y ax
is
Middle frequency
High frequency
minus15
minus1
minus05
minus5 minus4 minus3 minus2 minus1
120596 = 102 (rads)
120596 = 30 (rads)
Figure 7 Nyquist plot (magnification)
0
100
200
Gai
n (d
B)
0
90
Phas
e (de
g)
Frequency (rads)
minus300
minus200
minus100
minus270
minus180
minus90
10minus2 100 102 104
P(s) + F(s)
L(s)
K(s)
Frequency (rads)10minus2 100 102 104
Figure 8 Bode plot
we used the PID controller whose gain parameters are thesame as those of the I-PD controller Since both feedbacksystems have the same open-loop transfer function theirfeedback properties must be the same provided that eachinput signal does not saturate The simulation results of thestep response are shown in Figure 9 where the upper andlower parts are the output and input signals respectivelyThe solid curves represent the responses given by the I-PDcontroller while the dashed curves represent those by the PIDcontroller One can see that the input signal given by the PIDcontroller is saturated while that by the I-PD controller is notsaturated and satisfies the limitation (3) The output signalgiven by the I-PD controller settles down to the desired valuewithout any overshoot
It should be noted that the gain parameters in thedesigned controller are not tuned with the I-PD structure
Journal of Applied Mathematics 7
0 1 2 3 4 50
005
01
015
Am
plitu
de (m
)
0 1 2 3 4 5
0102030
Time (s)
Ang
le (d
eg)
I-PDPID
minus30
minus20
minus10
Time (s)
Figure 9 Step response simulation results
0 05 1 15 2 25 3 35 4 45 50
005
01
015
Time (s)
Am
plitu
de (m
)
0 05 1 15 2 25 3 35 4 45 5
0102030
Am
plitu
de (r
ad)
minus30
minus20
minus10
Time (s)
I-PDPID
Figure 10 Step response experimental results
In our experience it is difficult not to saturate the inputlimitation for the PID structure using any design method
6 Experimental Results
This section evaluates the I-PD controller whose designis given in Section 4 through an experiment of the stepresponse The PID controller used in Section 4 was alsoevaluated for comparison The results of trajectory trackingcontrol by the I-PD controller were also evaluated
61 Step Response Experiment The results of the stepresponse experiment are shown in Figure 10 where thedescription of the figure is the same as that of Figure 9 In thisexperiment the influence of the time delay appeared and theinput signals were slightly larger than those obtained in thesimulations The rise and settling time results are howeveralmost the same as those obtained in the simulation
62 Trajectory Tracking Experiments We tested two kindsof trajectories for tracking control a square and a circular
8 Journal of Applied Mathematics
0 002 004
0
002
004
006
Reference
minus006minus006
minus004
minus004
minus002
minus002 0 002 004
0
002
004
006
minus006minus006
minus004
minus004
minus002
minus002
yb
(m)
xb (m)
I-PD controlyb
(m)
xb (m)
ReferenceI-PD control
Figure 11 Experimental results of tracking control
0 5 10 15 20 25 30 35 40 45 50
0005
01
02040
Time (s)
0 5 10 15 20 25 30 35 40 45 50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
minus01minus005
minus40minus20
02040
minus40minus20
0005
01
minus01minus005
Time (s)
Time (s)Time (s)
ReferenceI-PD control
Controller outputActuator output
xb
(m)
120572(d
eg)
yb
(m)
120573(d
eg)
Figure 12 Experimental results of square response
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0005
01
02040
minus01minus005
minus40minus20
02040
minus40minus20
0005
01
minus01minus005
Time (s)Time (s)
Time (s)Time (s)
xb
(m)
120572(d
eg)
yb
(m)
120573(d
eg)
ReferenceI-PD control
Controller outputActuator output
Figure 13 Experimental results of circle response
Journal of Applied Mathematics 9
trajectory The side of the square trajectory was 01 ([m])and the radius of the circular trajectory was 005 ([m]) Theresults are shown in Figure 11 where the left part shows atrajectory of the square trajectory tracking control and theright part shows a trajectory of the circular trajectory trackingcontrol The time histories of the ball and input angles areshown in Figures 12 and 13 In the square trajectory trackingcontrol experiment the responses were similar to those inthe step response experiment except for a slight vibrationSuch vibration phenomena are noticeable in the responsesof circular trajectory tracking control in particular the casewhen the input signal is relatively small The reason for thesephenomena could be the friction of the ball against the plateor a backlash of the gear system
7 Conclusions
This paper applied the GKYP lemma to an open-looptransfer function including an I-PD controller and a noisereduction filter for the ball and plate systemThemultiple FDIspecifications for the finite frequency ranges were satisfiedby a solution of the LMI optimization problem The solutionincludes the optimal parameters in the I-PD controller Thefirst-order low-pass filter reduced the noise in the high fre-quency range and improved the steady-state response Bothsimulations and experiments evaluated the effectiveness ofthe designmethod by comparing the standard PID controller
The PI-D (proportional integral-derivative) control sys-tem which moved the derivative controller to the innerfeedback loop also has the same open-loop transfer functionas the standard PID controller Thus the approach in thispaper can also be applied to the PI-D controller
References
[1] MMoarref M Saadat and G Vossoughi ldquoMechatronic designand position control of a novel ball and plate systemrdquo inProceedings of the Mediterranean Conference on Control andAutomation (MEDrsquo 08) pp 1071ndash1076 June 2008
[2] P Kokotovic ldquoThe joy of feedback nonlinear and adaptiverdquoIEEE Control Systems Magazine vol 12 no 3 pp 7ndash17 1992
[3] F Zheng X Li X Qian and S Wang ldquoModeling and PIDneural network research for the ball and plate systemrdquo inProceedings of the International Conference on Electronics Com-munications and Control (ICECC rsquo11) pp 331ndash334 September2011
[4] X Fan N Zhang and S Teng ldquoTrajectory planning andtracking of ball andplate systemusing hierarchical fuzzy controlschemerdquo Fuzzy Sets and Systems vol 144 no 2 pp 297ndash3122004
[5] M Bai Y Tian and Y Wang ldquoDecoupled fuzzy sliding modecontrol to ball and plate systemrdquo in Proceedings of the 2ndInternational Conference on Intelligent Control and InformationProcessing (ICICIP rsquo11) pp 685ndash690 July 2011
[6] F Borrelli Constrained Optimal Control of Linear and HybridSystems vol 290 of Lecture Notes in Control and InformationSciences Springer 2003
[7] K J Astrom and T Hagglund Advanced PID Control Instru-mentation Systems 2005
[8] Q Li and Z Kemin Introduction to Feedback Control PrenticeHall 2009
[9] T Iwasaki and S Hara ldquoGeneralized KYP lemma unifiedfrequency domain inequalities with design applicationsrdquo IEEETransactions on Automatic Control vol 50 no 1 pp 41ndash592005
[10] A Rantzer ldquoOn the Kalman-Yakubovich-Popov lemmardquo Sys-tems amp Control Letters vol 28 no 1 pp 7ndash10 1996
[11] J Lofberg ldquoYALMIP a toolbox for modeling and optimizationin MATLABrdquo in Proceedings of the IEEE International Sympo-sium on Computer-Aided Control System Design (CACSD rsquo11)Taipei Taiwan 2004
[12] K C TohM J Todd and RH Tutuncu ldquoSDPT3mdashaMATLABsoftware package for semidefinite programming version 13rdquoOptimization Methods and Software vol 11 no 1 pp 545ndash5811999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 5
LPF
0 1 2 3 4 5 6 7 8Time (s)
Am
plitu
de (m
)
002040608
112141618
2
Frequency (rads)
Pow
er
minus49
minus485
minus48
minus475
times10minus3
100 101 102
Raw data
Figure 5 Time history and FFT analysis result of measurementsignal with low-pass filter
the same experimental setup as given in Section 22 Figure 5shows the spectral analysis results of the measurement andfiltered signals which are represented by the dotted and solidcurves respectively From these results it can be seen that thenoise at frequencies over 20 ([rads]) has been reduced
42 State-Space Realization of Open-Loop Transfer FunctionTo obtain an LMI based on the GKYP lemma a state-spacerealization of 119871(119904) is required to be affine with respect to aset of the design parameters 120588 = (119870
119901 119870119894 119870119889) If we fix 120591 in
119870(119904) at 10times10minus2 the design parameters appear affinely in thenumerator of119870(119904) Indeed the controllable canonical form of119870(119904) is written as
[119860119896119861119896(120588)
119862119896119863119896(120588)
] =
[[[[[[
[
0 0119870119894
120591
1 minus1
120591119870119894minus119870119889
1205912
0 1 119870119901+119870119889
120591
]]]]]]
]
(27)
Realizations of 119875(119904) and 119865(119904) are also written as
[119860119901
119861119901
119862119901119863119901
] = [
[
0 1 0
0 0 119870bap1 0 0
]
]
(28)
[119860119891
119861119891
119862119891
119863119891
] = [minus120583 1
120583 0] (29)
respectively By combining these realizations (27)ndash(29) weobtain a realization of 119871(119904) as
119871 (119904) = [119860 119861 (120588)
119862 119863 (120588)] (30)
where
119860 = [
[
119860119896
0 0
119862119896119861119901
119860119901
0
119862119896119863119901119861119891119862119901119861119891119860119891
]
]
119861 = [
[
119861119896(120588)
119863119896(120588) 119861119901
119863119896(120588)119863119901119861119891
]
]
119862 = [119862119896119863119901119861119891 119862119901119861119891119862119891] 119863 = 119863
119896(120588)119863119901119863119891
(31)
Consequently the state-space realization of 119871(119904) is affine withrespect to 120588
43 Specifications To shape the Nyquist plot of 119871(119904) werequire the following FDI specifications
minus2R [119871 (119895120596)] +I [119871 (119895120596)] gt 120574ℓforall120596 le 08 (32)
I [119871 (119895120596)] lt 120574119898 25 le
forall120596 le 28 (33)
1003816100381610038161003816119871 (119895120596)1003816100381610038161003816 lt 120574ℎ
forall120596 ge 10 (34)
Specification (32) with a large 120574ℓ(gt 0) ensures sensitivity
reduction in the low frequency range by making the gain of119871(119904) high Specification (33) requires the Nyquist plot to beoutside a circle with its center at the point minus1 + 1198950 so thata certain stability margin is guaranteed Specification (34)with a small 120574
ℎensures robustness against the unmodeled
dynamics that typically exists in the high frequency rangeIn addition to the above basic specifications we also
require the following FDIs that improve the property oftrajectory tracking
4R [119871 (119895120596)] +I [119871 (119895120596)] lt 1205741forall120596 ge 03 (35)
4R [119871 (119895120596)] +I [119871 (119895120596)] gt 1205742forall120596 ge 03 (36)
Since the integral action aloneworks on the error between theoutput and the reference signals the property of trajectorytracking depends directly on the integrator Here we focuson the corner angular frequency 120596
119868by the integral action
in the I-PD controller which is given by 120596119868= 1119879
119894 where
119879119894= 119870119901119870119894 We have a strong integral action and the error is
corrected quickly when the corner angular frequency is highwhile too high a corner angular frequency causes overshootand hunting Thus we impose restrictions for the phase ofthe I-PD controller It should be noted that the phase atlower frequencies is about minus90 ([deg]) while the phase atthe corner angular frequency is about 0 [deg] If we find thefrequency at a specific phase from minus90 ([deg]) to 0 ([deg])the corner angular frequency is greater than that frequencySpecifications (35) and (36) restrict the phase of the I-PDcontroller as well as the open-loop transfer function so thatthe corner angular frequency is greater than the frequency atthe lowest point in the frequency range 03 ([rads])
44 I-PD Controller Design We design an I-PD controller bymaximizing 120574
ℓsubject to Specifications (32)ndash(36) where
(120574119898 120574ℎ 1205741 1205742) = (minus1 05 1 minus100) (37)
6 Journal of Applied Mathematics
0 50 100 150 200
0
50
100
150
Real axis
Imag
inar
y ax
is
The upper limit
Low frequency
The lower limit
minus150
minus100
minus50
minus150minus200 minus100 minus50
Figure 6 Nyquist plot
That is the optimization problem is
max119870119894 119870119901119870119889 120574ℓ
120574ℓ
subject to (19)ndash(36) and (37)
(38)
Each of the design Specifications (32)ndash(36) is reduced to anLMI condition through Lemma 1 with the realization (30)The Specification (32) is modified to 120576 le forall120596 le 08 where 120576 =10 times 10
minus4 because 119871(119904) includes the origin poles that preventus from taking 120596 = 0 Then the LMI optimization problemis to maximize 120574
ℓsubject to all these LMI conditions where
119870119894 119870119901 and 119870
119889are the common decision variables while
1198751 119875
5and 119876
1 119876
5appear in the LMIs as independent
decision variables It should be noted that 120574ℓappears alone in
Π22in the LMI condition (25) corresponding to (36) In this
sense 120574ℓis also an independent decision variable It should
also be noted that 119875 and 119876 in (25) appear in each of the LMIconditions as the independent decision variables
To solve this LMI optimization problem we use YALMIPR20120806 [11] an LMI parser and SPDT3 version 40 [12]an LMI solver on MATLAB R2011b The resulting optimalparameters in the I-PD controller and 120574
ℓare
(119870119901 119870119894 119870119889 120574ℓ) = (92859 107806 4138 119464)
(39)
The Nyquist plots are shown in Figures 6 and 7 where 119871(119895120596)satisfies the design specifications given in Section 43 Since 120574
ℓ
is maximized sensitivity is reduced in the frequency rangeThe Bode plots of 119865(119904)119875(119904) 119871(119904) and 119870(119904) are shown inFigure 8 where the corner angular frequency 120596
119868of the I-PD
controller is larger than 03 ([rads])
5 Simulation Results
The I-PD controller whose design is described in the previoussection was evaluated by a simulation of the step response Tocompare the response with that of a standard PID controller
0 1
0
05
1
15
Real axis
Imag
inar
y ax
is
Middle frequency
High frequency
minus15
minus1
minus05
minus5 minus4 minus3 minus2 minus1
120596 = 102 (rads)
120596 = 30 (rads)
Figure 7 Nyquist plot (magnification)
0
100
200
Gai
n (d
B)
0
90
Phas
e (de
g)
Frequency (rads)
minus300
minus200
minus100
minus270
minus180
minus90
10minus2 100 102 104
P(s) + F(s)
L(s)
K(s)
Frequency (rads)10minus2 100 102 104
Figure 8 Bode plot
we used the PID controller whose gain parameters are thesame as those of the I-PD controller Since both feedbacksystems have the same open-loop transfer function theirfeedback properties must be the same provided that eachinput signal does not saturate The simulation results of thestep response are shown in Figure 9 where the upper andlower parts are the output and input signals respectivelyThe solid curves represent the responses given by the I-PDcontroller while the dashed curves represent those by the PIDcontroller One can see that the input signal given by the PIDcontroller is saturated while that by the I-PD controller is notsaturated and satisfies the limitation (3) The output signalgiven by the I-PD controller settles down to the desired valuewithout any overshoot
It should be noted that the gain parameters in thedesigned controller are not tuned with the I-PD structure
Journal of Applied Mathematics 7
0 1 2 3 4 50
005
01
015
Am
plitu
de (m
)
0 1 2 3 4 5
0102030
Time (s)
Ang
le (d
eg)
I-PDPID
minus30
minus20
minus10
Time (s)
Figure 9 Step response simulation results
0 05 1 15 2 25 3 35 4 45 50
005
01
015
Time (s)
Am
plitu
de (m
)
0 05 1 15 2 25 3 35 4 45 5
0102030
Am
plitu
de (r
ad)
minus30
minus20
minus10
Time (s)
I-PDPID
Figure 10 Step response experimental results
In our experience it is difficult not to saturate the inputlimitation for the PID structure using any design method
6 Experimental Results
This section evaluates the I-PD controller whose designis given in Section 4 through an experiment of the stepresponse The PID controller used in Section 4 was alsoevaluated for comparison The results of trajectory trackingcontrol by the I-PD controller were also evaluated
61 Step Response Experiment The results of the stepresponse experiment are shown in Figure 10 where thedescription of the figure is the same as that of Figure 9 In thisexperiment the influence of the time delay appeared and theinput signals were slightly larger than those obtained in thesimulations The rise and settling time results are howeveralmost the same as those obtained in the simulation
62 Trajectory Tracking Experiments We tested two kindsof trajectories for tracking control a square and a circular
8 Journal of Applied Mathematics
0 002 004
0
002
004
006
Reference
minus006minus006
minus004
minus004
minus002
minus002 0 002 004
0
002
004
006
minus006minus006
minus004
minus004
minus002
minus002
yb
(m)
xb (m)
I-PD controlyb
(m)
xb (m)
ReferenceI-PD control
Figure 11 Experimental results of tracking control
0 5 10 15 20 25 30 35 40 45 50
0005
01
02040
Time (s)
0 5 10 15 20 25 30 35 40 45 50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
minus01minus005
minus40minus20
02040
minus40minus20
0005
01
minus01minus005
Time (s)
Time (s)Time (s)
ReferenceI-PD control
Controller outputActuator output
xb
(m)
120572(d
eg)
yb
(m)
120573(d
eg)
Figure 12 Experimental results of square response
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0005
01
02040
minus01minus005
minus40minus20
02040
minus40minus20
0005
01
minus01minus005
Time (s)Time (s)
Time (s)Time (s)
xb
(m)
120572(d
eg)
yb
(m)
120573(d
eg)
ReferenceI-PD control
Controller outputActuator output
Figure 13 Experimental results of circle response
Journal of Applied Mathematics 9
trajectory The side of the square trajectory was 01 ([m])and the radius of the circular trajectory was 005 ([m]) Theresults are shown in Figure 11 where the left part shows atrajectory of the square trajectory tracking control and theright part shows a trajectory of the circular trajectory trackingcontrol The time histories of the ball and input angles areshown in Figures 12 and 13 In the square trajectory trackingcontrol experiment the responses were similar to those inthe step response experiment except for a slight vibrationSuch vibration phenomena are noticeable in the responsesof circular trajectory tracking control in particular the casewhen the input signal is relatively small The reason for thesephenomena could be the friction of the ball against the plateor a backlash of the gear system
7 Conclusions
This paper applied the GKYP lemma to an open-looptransfer function including an I-PD controller and a noisereduction filter for the ball and plate systemThemultiple FDIspecifications for the finite frequency ranges were satisfiedby a solution of the LMI optimization problem The solutionincludes the optimal parameters in the I-PD controller Thefirst-order low-pass filter reduced the noise in the high fre-quency range and improved the steady-state response Bothsimulations and experiments evaluated the effectiveness ofthe designmethod by comparing the standard PID controller
The PI-D (proportional integral-derivative) control sys-tem which moved the derivative controller to the innerfeedback loop also has the same open-loop transfer functionas the standard PID controller Thus the approach in thispaper can also be applied to the PI-D controller
References
[1] MMoarref M Saadat and G Vossoughi ldquoMechatronic designand position control of a novel ball and plate systemrdquo inProceedings of the Mediterranean Conference on Control andAutomation (MEDrsquo 08) pp 1071ndash1076 June 2008
[2] P Kokotovic ldquoThe joy of feedback nonlinear and adaptiverdquoIEEE Control Systems Magazine vol 12 no 3 pp 7ndash17 1992
[3] F Zheng X Li X Qian and S Wang ldquoModeling and PIDneural network research for the ball and plate systemrdquo inProceedings of the International Conference on Electronics Com-munications and Control (ICECC rsquo11) pp 331ndash334 September2011
[4] X Fan N Zhang and S Teng ldquoTrajectory planning andtracking of ball andplate systemusing hierarchical fuzzy controlschemerdquo Fuzzy Sets and Systems vol 144 no 2 pp 297ndash3122004
[5] M Bai Y Tian and Y Wang ldquoDecoupled fuzzy sliding modecontrol to ball and plate systemrdquo in Proceedings of the 2ndInternational Conference on Intelligent Control and InformationProcessing (ICICIP rsquo11) pp 685ndash690 July 2011
[6] F Borrelli Constrained Optimal Control of Linear and HybridSystems vol 290 of Lecture Notes in Control and InformationSciences Springer 2003
[7] K J Astrom and T Hagglund Advanced PID Control Instru-mentation Systems 2005
[8] Q Li and Z Kemin Introduction to Feedback Control PrenticeHall 2009
[9] T Iwasaki and S Hara ldquoGeneralized KYP lemma unifiedfrequency domain inequalities with design applicationsrdquo IEEETransactions on Automatic Control vol 50 no 1 pp 41ndash592005
[10] A Rantzer ldquoOn the Kalman-Yakubovich-Popov lemmardquo Sys-tems amp Control Letters vol 28 no 1 pp 7ndash10 1996
[11] J Lofberg ldquoYALMIP a toolbox for modeling and optimizationin MATLABrdquo in Proceedings of the IEEE International Sympo-sium on Computer-Aided Control System Design (CACSD rsquo11)Taipei Taiwan 2004
[12] K C TohM J Todd and RH Tutuncu ldquoSDPT3mdashaMATLABsoftware package for semidefinite programming version 13rdquoOptimization Methods and Software vol 11 no 1 pp 545ndash5811999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
6 Journal of Applied Mathematics
0 50 100 150 200
0
50
100
150
Real axis
Imag
inar
y ax
is
The upper limit
Low frequency
The lower limit
minus150
minus100
minus50
minus150minus200 minus100 minus50
Figure 6 Nyquist plot
That is the optimization problem is
max119870119894 119870119901119870119889 120574ℓ
120574ℓ
subject to (19)ndash(36) and (37)
(38)
Each of the design Specifications (32)ndash(36) is reduced to anLMI condition through Lemma 1 with the realization (30)The Specification (32) is modified to 120576 le forall120596 le 08 where 120576 =10 times 10
minus4 because 119871(119904) includes the origin poles that preventus from taking 120596 = 0 Then the LMI optimization problemis to maximize 120574
ℓsubject to all these LMI conditions where
119870119894 119870119901 and 119870
119889are the common decision variables while
1198751 119875
5and 119876
1 119876
5appear in the LMIs as independent
decision variables It should be noted that 120574ℓappears alone in
Π22in the LMI condition (25) corresponding to (36) In this
sense 120574ℓis also an independent decision variable It should
also be noted that 119875 and 119876 in (25) appear in each of the LMIconditions as the independent decision variables
To solve this LMI optimization problem we use YALMIPR20120806 [11] an LMI parser and SPDT3 version 40 [12]an LMI solver on MATLAB R2011b The resulting optimalparameters in the I-PD controller and 120574
ℓare
(119870119901 119870119894 119870119889 120574ℓ) = (92859 107806 4138 119464)
(39)
The Nyquist plots are shown in Figures 6 and 7 where 119871(119895120596)satisfies the design specifications given in Section 43 Since 120574
ℓ
is maximized sensitivity is reduced in the frequency rangeThe Bode plots of 119865(119904)119875(119904) 119871(119904) and 119870(119904) are shown inFigure 8 where the corner angular frequency 120596
119868of the I-PD
controller is larger than 03 ([rads])
5 Simulation Results
The I-PD controller whose design is described in the previoussection was evaluated by a simulation of the step response Tocompare the response with that of a standard PID controller
0 1
0
05
1
15
Real axis
Imag
inar
y ax
is
Middle frequency
High frequency
minus15
minus1
minus05
minus5 minus4 minus3 minus2 minus1
120596 = 102 (rads)
120596 = 30 (rads)
Figure 7 Nyquist plot (magnification)
0
100
200
Gai
n (d
B)
0
90
Phas
e (de
g)
Frequency (rads)
minus300
minus200
minus100
minus270
minus180
minus90
10minus2 100 102 104
P(s) + F(s)
L(s)
K(s)
Frequency (rads)10minus2 100 102 104
Figure 8 Bode plot
we used the PID controller whose gain parameters are thesame as those of the I-PD controller Since both feedbacksystems have the same open-loop transfer function theirfeedback properties must be the same provided that eachinput signal does not saturate The simulation results of thestep response are shown in Figure 9 where the upper andlower parts are the output and input signals respectivelyThe solid curves represent the responses given by the I-PDcontroller while the dashed curves represent those by the PIDcontroller One can see that the input signal given by the PIDcontroller is saturated while that by the I-PD controller is notsaturated and satisfies the limitation (3) The output signalgiven by the I-PD controller settles down to the desired valuewithout any overshoot
It should be noted that the gain parameters in thedesigned controller are not tuned with the I-PD structure
Journal of Applied Mathematics 7
0 1 2 3 4 50
005
01
015
Am
plitu
de (m
)
0 1 2 3 4 5
0102030
Time (s)
Ang
le (d
eg)
I-PDPID
minus30
minus20
minus10
Time (s)
Figure 9 Step response simulation results
0 05 1 15 2 25 3 35 4 45 50
005
01
015
Time (s)
Am
plitu
de (m
)
0 05 1 15 2 25 3 35 4 45 5
0102030
Am
plitu
de (r
ad)
minus30
minus20
minus10
Time (s)
I-PDPID
Figure 10 Step response experimental results
In our experience it is difficult not to saturate the inputlimitation for the PID structure using any design method
6 Experimental Results
This section evaluates the I-PD controller whose designis given in Section 4 through an experiment of the stepresponse The PID controller used in Section 4 was alsoevaluated for comparison The results of trajectory trackingcontrol by the I-PD controller were also evaluated
61 Step Response Experiment The results of the stepresponse experiment are shown in Figure 10 where thedescription of the figure is the same as that of Figure 9 In thisexperiment the influence of the time delay appeared and theinput signals were slightly larger than those obtained in thesimulations The rise and settling time results are howeveralmost the same as those obtained in the simulation
62 Trajectory Tracking Experiments We tested two kindsof trajectories for tracking control a square and a circular
8 Journal of Applied Mathematics
0 002 004
0
002
004
006
Reference
minus006minus006
minus004
minus004
minus002
minus002 0 002 004
0
002
004
006
minus006minus006
minus004
minus004
minus002
minus002
yb
(m)
xb (m)
I-PD controlyb
(m)
xb (m)
ReferenceI-PD control
Figure 11 Experimental results of tracking control
0 5 10 15 20 25 30 35 40 45 50
0005
01
02040
Time (s)
0 5 10 15 20 25 30 35 40 45 50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
minus01minus005
minus40minus20
02040
minus40minus20
0005
01
minus01minus005
Time (s)
Time (s)Time (s)
ReferenceI-PD control
Controller outputActuator output
xb
(m)
120572(d
eg)
yb
(m)
120573(d
eg)
Figure 12 Experimental results of square response
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0005
01
02040
minus01minus005
minus40minus20
02040
minus40minus20
0005
01
minus01minus005
Time (s)Time (s)
Time (s)Time (s)
xb
(m)
120572(d
eg)
yb
(m)
120573(d
eg)
ReferenceI-PD control
Controller outputActuator output
Figure 13 Experimental results of circle response
Journal of Applied Mathematics 9
trajectory The side of the square trajectory was 01 ([m])and the radius of the circular trajectory was 005 ([m]) Theresults are shown in Figure 11 where the left part shows atrajectory of the square trajectory tracking control and theright part shows a trajectory of the circular trajectory trackingcontrol The time histories of the ball and input angles areshown in Figures 12 and 13 In the square trajectory trackingcontrol experiment the responses were similar to those inthe step response experiment except for a slight vibrationSuch vibration phenomena are noticeable in the responsesof circular trajectory tracking control in particular the casewhen the input signal is relatively small The reason for thesephenomena could be the friction of the ball against the plateor a backlash of the gear system
7 Conclusions
This paper applied the GKYP lemma to an open-looptransfer function including an I-PD controller and a noisereduction filter for the ball and plate systemThemultiple FDIspecifications for the finite frequency ranges were satisfiedby a solution of the LMI optimization problem The solutionincludes the optimal parameters in the I-PD controller Thefirst-order low-pass filter reduced the noise in the high fre-quency range and improved the steady-state response Bothsimulations and experiments evaluated the effectiveness ofthe designmethod by comparing the standard PID controller
The PI-D (proportional integral-derivative) control sys-tem which moved the derivative controller to the innerfeedback loop also has the same open-loop transfer functionas the standard PID controller Thus the approach in thispaper can also be applied to the PI-D controller
References
[1] MMoarref M Saadat and G Vossoughi ldquoMechatronic designand position control of a novel ball and plate systemrdquo inProceedings of the Mediterranean Conference on Control andAutomation (MEDrsquo 08) pp 1071ndash1076 June 2008
[2] P Kokotovic ldquoThe joy of feedback nonlinear and adaptiverdquoIEEE Control Systems Magazine vol 12 no 3 pp 7ndash17 1992
[3] F Zheng X Li X Qian and S Wang ldquoModeling and PIDneural network research for the ball and plate systemrdquo inProceedings of the International Conference on Electronics Com-munications and Control (ICECC rsquo11) pp 331ndash334 September2011
[4] X Fan N Zhang and S Teng ldquoTrajectory planning andtracking of ball andplate systemusing hierarchical fuzzy controlschemerdquo Fuzzy Sets and Systems vol 144 no 2 pp 297ndash3122004
[5] M Bai Y Tian and Y Wang ldquoDecoupled fuzzy sliding modecontrol to ball and plate systemrdquo in Proceedings of the 2ndInternational Conference on Intelligent Control and InformationProcessing (ICICIP rsquo11) pp 685ndash690 July 2011
[6] F Borrelli Constrained Optimal Control of Linear and HybridSystems vol 290 of Lecture Notes in Control and InformationSciences Springer 2003
[7] K J Astrom and T Hagglund Advanced PID Control Instru-mentation Systems 2005
[8] Q Li and Z Kemin Introduction to Feedback Control PrenticeHall 2009
[9] T Iwasaki and S Hara ldquoGeneralized KYP lemma unifiedfrequency domain inequalities with design applicationsrdquo IEEETransactions on Automatic Control vol 50 no 1 pp 41ndash592005
[10] A Rantzer ldquoOn the Kalman-Yakubovich-Popov lemmardquo Sys-tems amp Control Letters vol 28 no 1 pp 7ndash10 1996
[11] J Lofberg ldquoYALMIP a toolbox for modeling and optimizationin MATLABrdquo in Proceedings of the IEEE International Sympo-sium on Computer-Aided Control System Design (CACSD rsquo11)Taipei Taiwan 2004
[12] K C TohM J Todd and RH Tutuncu ldquoSDPT3mdashaMATLABsoftware package for semidefinite programming version 13rdquoOptimization Methods and Software vol 11 no 1 pp 545ndash5811999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 7
0 1 2 3 4 50
005
01
015
Am
plitu
de (m
)
0 1 2 3 4 5
0102030
Time (s)
Ang
le (d
eg)
I-PDPID
minus30
minus20
minus10
Time (s)
Figure 9 Step response simulation results
0 05 1 15 2 25 3 35 4 45 50
005
01
015
Time (s)
Am
plitu
de (m
)
0 05 1 15 2 25 3 35 4 45 5
0102030
Am
plitu
de (r
ad)
minus30
minus20
minus10
Time (s)
I-PDPID
Figure 10 Step response experimental results
In our experience it is difficult not to saturate the inputlimitation for the PID structure using any design method
6 Experimental Results
This section evaluates the I-PD controller whose designis given in Section 4 through an experiment of the stepresponse The PID controller used in Section 4 was alsoevaluated for comparison The results of trajectory trackingcontrol by the I-PD controller were also evaluated
61 Step Response Experiment The results of the stepresponse experiment are shown in Figure 10 where thedescription of the figure is the same as that of Figure 9 In thisexperiment the influence of the time delay appeared and theinput signals were slightly larger than those obtained in thesimulations The rise and settling time results are howeveralmost the same as those obtained in the simulation
62 Trajectory Tracking Experiments We tested two kindsof trajectories for tracking control a square and a circular
8 Journal of Applied Mathematics
0 002 004
0
002
004
006
Reference
minus006minus006
minus004
minus004
minus002
minus002 0 002 004
0
002
004
006
minus006minus006
minus004
minus004
minus002
minus002
yb
(m)
xb (m)
I-PD controlyb
(m)
xb (m)
ReferenceI-PD control
Figure 11 Experimental results of tracking control
0 5 10 15 20 25 30 35 40 45 50
0005
01
02040
Time (s)
0 5 10 15 20 25 30 35 40 45 50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
minus01minus005
minus40minus20
02040
minus40minus20
0005
01
minus01minus005
Time (s)
Time (s)Time (s)
ReferenceI-PD control
Controller outputActuator output
xb
(m)
120572(d
eg)
yb
(m)
120573(d
eg)
Figure 12 Experimental results of square response
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0005
01
02040
minus01minus005
minus40minus20
02040
minus40minus20
0005
01
minus01minus005
Time (s)Time (s)
Time (s)Time (s)
xb
(m)
120572(d
eg)
yb
(m)
120573(d
eg)
ReferenceI-PD control
Controller outputActuator output
Figure 13 Experimental results of circle response
Journal of Applied Mathematics 9
trajectory The side of the square trajectory was 01 ([m])and the radius of the circular trajectory was 005 ([m]) Theresults are shown in Figure 11 where the left part shows atrajectory of the square trajectory tracking control and theright part shows a trajectory of the circular trajectory trackingcontrol The time histories of the ball and input angles areshown in Figures 12 and 13 In the square trajectory trackingcontrol experiment the responses were similar to those inthe step response experiment except for a slight vibrationSuch vibration phenomena are noticeable in the responsesof circular trajectory tracking control in particular the casewhen the input signal is relatively small The reason for thesephenomena could be the friction of the ball against the plateor a backlash of the gear system
7 Conclusions
This paper applied the GKYP lemma to an open-looptransfer function including an I-PD controller and a noisereduction filter for the ball and plate systemThemultiple FDIspecifications for the finite frequency ranges were satisfiedby a solution of the LMI optimization problem The solutionincludes the optimal parameters in the I-PD controller Thefirst-order low-pass filter reduced the noise in the high fre-quency range and improved the steady-state response Bothsimulations and experiments evaluated the effectiveness ofthe designmethod by comparing the standard PID controller
The PI-D (proportional integral-derivative) control sys-tem which moved the derivative controller to the innerfeedback loop also has the same open-loop transfer functionas the standard PID controller Thus the approach in thispaper can also be applied to the PI-D controller
References
[1] MMoarref M Saadat and G Vossoughi ldquoMechatronic designand position control of a novel ball and plate systemrdquo inProceedings of the Mediterranean Conference on Control andAutomation (MEDrsquo 08) pp 1071ndash1076 June 2008
[2] P Kokotovic ldquoThe joy of feedback nonlinear and adaptiverdquoIEEE Control Systems Magazine vol 12 no 3 pp 7ndash17 1992
[3] F Zheng X Li X Qian and S Wang ldquoModeling and PIDneural network research for the ball and plate systemrdquo inProceedings of the International Conference on Electronics Com-munications and Control (ICECC rsquo11) pp 331ndash334 September2011
[4] X Fan N Zhang and S Teng ldquoTrajectory planning andtracking of ball andplate systemusing hierarchical fuzzy controlschemerdquo Fuzzy Sets and Systems vol 144 no 2 pp 297ndash3122004
[5] M Bai Y Tian and Y Wang ldquoDecoupled fuzzy sliding modecontrol to ball and plate systemrdquo in Proceedings of the 2ndInternational Conference on Intelligent Control and InformationProcessing (ICICIP rsquo11) pp 685ndash690 July 2011
[6] F Borrelli Constrained Optimal Control of Linear and HybridSystems vol 290 of Lecture Notes in Control and InformationSciences Springer 2003
[7] K J Astrom and T Hagglund Advanced PID Control Instru-mentation Systems 2005
[8] Q Li and Z Kemin Introduction to Feedback Control PrenticeHall 2009
[9] T Iwasaki and S Hara ldquoGeneralized KYP lemma unifiedfrequency domain inequalities with design applicationsrdquo IEEETransactions on Automatic Control vol 50 no 1 pp 41ndash592005
[10] A Rantzer ldquoOn the Kalman-Yakubovich-Popov lemmardquo Sys-tems amp Control Letters vol 28 no 1 pp 7ndash10 1996
[11] J Lofberg ldquoYALMIP a toolbox for modeling and optimizationin MATLABrdquo in Proceedings of the IEEE International Sympo-sium on Computer-Aided Control System Design (CACSD rsquo11)Taipei Taiwan 2004
[12] K C TohM J Todd and RH Tutuncu ldquoSDPT3mdashaMATLABsoftware package for semidefinite programming version 13rdquoOptimization Methods and Software vol 11 no 1 pp 545ndash5811999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
8 Journal of Applied Mathematics
0 002 004
0
002
004
006
Reference
minus006minus006
minus004
minus004
minus002
minus002 0 002 004
0
002
004
006
minus006minus006
minus004
minus004
minus002
minus002
yb
(m)
xb (m)
I-PD controlyb
(m)
xb (m)
ReferenceI-PD control
Figure 11 Experimental results of tracking control
0 5 10 15 20 25 30 35 40 45 50
0005
01
02040
Time (s)
0 5 10 15 20 25 30 35 40 45 50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
minus01minus005
minus40minus20
02040
minus40minus20
0005
01
minus01minus005
Time (s)
Time (s)Time (s)
ReferenceI-PD control
Controller outputActuator output
xb
(m)
120572(d
eg)
yb
(m)
120573(d
eg)
Figure 12 Experimental results of square response
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
0005
01
02040
minus01minus005
minus40minus20
02040
minus40minus20
0005
01
minus01minus005
Time (s)Time (s)
Time (s)Time (s)
xb
(m)
120572(d
eg)
yb
(m)
120573(d
eg)
ReferenceI-PD control
Controller outputActuator output
Figure 13 Experimental results of circle response
Journal of Applied Mathematics 9
trajectory The side of the square trajectory was 01 ([m])and the radius of the circular trajectory was 005 ([m]) Theresults are shown in Figure 11 where the left part shows atrajectory of the square trajectory tracking control and theright part shows a trajectory of the circular trajectory trackingcontrol The time histories of the ball and input angles areshown in Figures 12 and 13 In the square trajectory trackingcontrol experiment the responses were similar to those inthe step response experiment except for a slight vibrationSuch vibration phenomena are noticeable in the responsesof circular trajectory tracking control in particular the casewhen the input signal is relatively small The reason for thesephenomena could be the friction of the ball against the plateor a backlash of the gear system
7 Conclusions
This paper applied the GKYP lemma to an open-looptransfer function including an I-PD controller and a noisereduction filter for the ball and plate systemThemultiple FDIspecifications for the finite frequency ranges were satisfiedby a solution of the LMI optimization problem The solutionincludes the optimal parameters in the I-PD controller Thefirst-order low-pass filter reduced the noise in the high fre-quency range and improved the steady-state response Bothsimulations and experiments evaluated the effectiveness ofthe designmethod by comparing the standard PID controller
The PI-D (proportional integral-derivative) control sys-tem which moved the derivative controller to the innerfeedback loop also has the same open-loop transfer functionas the standard PID controller Thus the approach in thispaper can also be applied to the PI-D controller
References
[1] MMoarref M Saadat and G Vossoughi ldquoMechatronic designand position control of a novel ball and plate systemrdquo inProceedings of the Mediterranean Conference on Control andAutomation (MEDrsquo 08) pp 1071ndash1076 June 2008
[2] P Kokotovic ldquoThe joy of feedback nonlinear and adaptiverdquoIEEE Control Systems Magazine vol 12 no 3 pp 7ndash17 1992
[3] F Zheng X Li X Qian and S Wang ldquoModeling and PIDneural network research for the ball and plate systemrdquo inProceedings of the International Conference on Electronics Com-munications and Control (ICECC rsquo11) pp 331ndash334 September2011
[4] X Fan N Zhang and S Teng ldquoTrajectory planning andtracking of ball andplate systemusing hierarchical fuzzy controlschemerdquo Fuzzy Sets and Systems vol 144 no 2 pp 297ndash3122004
[5] M Bai Y Tian and Y Wang ldquoDecoupled fuzzy sliding modecontrol to ball and plate systemrdquo in Proceedings of the 2ndInternational Conference on Intelligent Control and InformationProcessing (ICICIP rsquo11) pp 685ndash690 July 2011
[6] F Borrelli Constrained Optimal Control of Linear and HybridSystems vol 290 of Lecture Notes in Control and InformationSciences Springer 2003
[7] K J Astrom and T Hagglund Advanced PID Control Instru-mentation Systems 2005
[8] Q Li and Z Kemin Introduction to Feedback Control PrenticeHall 2009
[9] T Iwasaki and S Hara ldquoGeneralized KYP lemma unifiedfrequency domain inequalities with design applicationsrdquo IEEETransactions on Automatic Control vol 50 no 1 pp 41ndash592005
[10] A Rantzer ldquoOn the Kalman-Yakubovich-Popov lemmardquo Sys-tems amp Control Letters vol 28 no 1 pp 7ndash10 1996
[11] J Lofberg ldquoYALMIP a toolbox for modeling and optimizationin MATLABrdquo in Proceedings of the IEEE International Sympo-sium on Computer-Aided Control System Design (CACSD rsquo11)Taipei Taiwan 2004
[12] K C TohM J Todd and RH Tutuncu ldquoSDPT3mdashaMATLABsoftware package for semidefinite programming version 13rdquoOptimization Methods and Software vol 11 no 1 pp 545ndash5811999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 9
trajectory The side of the square trajectory was 01 ([m])and the radius of the circular trajectory was 005 ([m]) Theresults are shown in Figure 11 where the left part shows atrajectory of the square trajectory tracking control and theright part shows a trajectory of the circular trajectory trackingcontrol The time histories of the ball and input angles areshown in Figures 12 and 13 In the square trajectory trackingcontrol experiment the responses were similar to those inthe step response experiment except for a slight vibrationSuch vibration phenomena are noticeable in the responsesof circular trajectory tracking control in particular the casewhen the input signal is relatively small The reason for thesephenomena could be the friction of the ball against the plateor a backlash of the gear system
7 Conclusions
This paper applied the GKYP lemma to an open-looptransfer function including an I-PD controller and a noisereduction filter for the ball and plate systemThemultiple FDIspecifications for the finite frequency ranges were satisfiedby a solution of the LMI optimization problem The solutionincludes the optimal parameters in the I-PD controller Thefirst-order low-pass filter reduced the noise in the high fre-quency range and improved the steady-state response Bothsimulations and experiments evaluated the effectiveness ofthe designmethod by comparing the standard PID controller
The PI-D (proportional integral-derivative) control sys-tem which moved the derivative controller to the innerfeedback loop also has the same open-loop transfer functionas the standard PID controller Thus the approach in thispaper can also be applied to the PI-D controller
References
[1] MMoarref M Saadat and G Vossoughi ldquoMechatronic designand position control of a novel ball and plate systemrdquo inProceedings of the Mediterranean Conference on Control andAutomation (MEDrsquo 08) pp 1071ndash1076 June 2008
[2] P Kokotovic ldquoThe joy of feedback nonlinear and adaptiverdquoIEEE Control Systems Magazine vol 12 no 3 pp 7ndash17 1992
[3] F Zheng X Li X Qian and S Wang ldquoModeling and PIDneural network research for the ball and plate systemrdquo inProceedings of the International Conference on Electronics Com-munications and Control (ICECC rsquo11) pp 331ndash334 September2011
[4] X Fan N Zhang and S Teng ldquoTrajectory planning andtracking of ball andplate systemusing hierarchical fuzzy controlschemerdquo Fuzzy Sets and Systems vol 144 no 2 pp 297ndash3122004
[5] M Bai Y Tian and Y Wang ldquoDecoupled fuzzy sliding modecontrol to ball and plate systemrdquo in Proceedings of the 2ndInternational Conference on Intelligent Control and InformationProcessing (ICICIP rsquo11) pp 685ndash690 July 2011
[6] F Borrelli Constrained Optimal Control of Linear and HybridSystems vol 290 of Lecture Notes in Control and InformationSciences Springer 2003
[7] K J Astrom and T Hagglund Advanced PID Control Instru-mentation Systems 2005
[8] Q Li and Z Kemin Introduction to Feedback Control PrenticeHall 2009
[9] T Iwasaki and S Hara ldquoGeneralized KYP lemma unifiedfrequency domain inequalities with design applicationsrdquo IEEETransactions on Automatic Control vol 50 no 1 pp 41ndash592005
[10] A Rantzer ldquoOn the Kalman-Yakubovich-Popov lemmardquo Sys-tems amp Control Letters vol 28 no 1 pp 7ndash10 1996
[11] J Lofberg ldquoYALMIP a toolbox for modeling and optimizationin MATLABrdquo in Proceedings of the IEEE International Sympo-sium on Computer-Aided Control System Design (CACSD rsquo11)Taipei Taiwan 2004
[12] K C TohM J Todd and RH Tutuncu ldquoSDPT3mdashaMATLABsoftware package for semidefinite programming version 13rdquoOptimization Methods and Software vol 11 no 1 pp 545ndash5811999
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of