research article tracking control of a leg rehabilitation...
TRANSCRIPT
Research ArticleTracking Control of a Leg Rehabilitation MachineDriven by Pneumatic Artificial Muscles Using CompositeFuzzy Theory
Ming-Kun Chang
Department of Mechanical and Computer-Aided Engineering St Johnrsquos University No 499 Section 4 Tam King RoadTamsui District New Taipei City 25135 Taiwan
Correspondence should be addressed to Ming-Kun Chang mkchangmailsjuedutw
Received 14 November 2013 Accepted 22 December 2013 Published 18 March 2014
Academic Editors N-I Kim and K I Ramachandran
Copyright copy 2014 Ming-Kun Chang This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
It is difficult to achieve excellent tracking performance for a two-joint leg rehabilitation machine driven by pneumatic artificialmuscles (PAMs) because the system has a coupling effect highly nonlinear and time-varying behavior associated with gascompression and the nonlinear elasticity of bladder containers This paper therefore proposes a T-S fuzzy theory with supervisorycontrol in order to overcome the above problems The T-S fuzzy theory decomposes the model of a nonlinear system into a set oflinear subsystems In this manner the controller in the T-S fuzzy model is able to use simple linear control techniques to providea systematic framework for the design of a state feedback controller Then the LMI Toolbox of MATLAB can be employed tosolve linear matrix inequalities (LMIs) in order to determine controller gains based on the Lyapunov direct method Moreover thesupervisory control can overcome the coupling effect for a leg rehabilitation machine Experimental results show that the proposedcontroller can achieve excellent tracking performance and guarantee robustness to system parameter uncertainties
1 Introduction
In cases of traumatic brain injury bone injury amputationor spinal cord injury caused by misfortunes such as trafficaccidents and cerebral apoplexy lower limb rehabilitationmachine can help patients recover extremity functions bymeans of continuous passive motion (CPM) Traditionallyphysical therapy for achieving functional rehabilitation iscarried out bymedical therapists on a person-to-person basisHowever recently many automatic rehabilitation deviceshave been gradually applied in physical therapy programsRehabilitationmachines are usually driven by electricmotorswhich are typically rigid in nature Because of this actuatorscan generate discomfort or pain when interfacing withhumans For this reason current electromechanical actuationsystems should be replaced to ensure adaptability conformityand safety An adequate actuator for a rehabilitation devicemust provide physically adjustable compliance and safety andensure soft contact with the patient similar to the behaviorof human muscles It has been suggested that pneumaticartificial muscles (PAMs) can contribute towards achieving
more comfortable devices for interfacing with human limbsegments
PAMs behave in amanner very similar to themuscles thatmove the skeletons of animals and have many advantagessuch as high power to weight ratio [1] high power to volumeratio [2] low maintenance negligible mechanical wear lowcost cleanliness high reliability flexibility and compliancefor use with humans For these reasons PAMs are commonlyemployed in rehabilitation engineering nursing and human-friendly therapeutic machine
However PAMs exhibit highly nonlinear and time-vary-ing behavior due to the compression of air and the nonlinearelasticity of bladder containers This makes it difficult forclassical controllers to achieve excellent control performanceIn recent years researchers have developed a wide varietyof approaches to overcome these problems Noritsugu andTanaka [3] developed four modes of linear motion withimpedance control to control force during movement andused an adaptive identificationmethod to estimate the systemmodel Lilly and Yang [4] applied a slidingmode controller toa planar arm actuated by two PMA groups simulation results
Hindawi Publishing Corporatione Scientific World JournalVolume 2014 Article ID 464276 12 pageshttpdxdoiorg1011552014464276
2 The Scientific World Journal
were consistent with theoretical findings for two differentmasses Ahn and Anh [5] adopted an ARNN controller ina PAM manipulator for reducing tracking errors Shen [6]developed a full nonlinear model that encompassed allthe major existing nonlinearities Based on this model thestandard slidingmode control approachwas applied to obtainrobust control even in the event of model uncertainties anddisturbances
Since the inception of fuzzy set theory by Zadeh [7] in1965 a great deal of research has been focused on fuzzycontrol systems Takagi and Sugeno [8] proposed the T-S fuzzy model-based controller in 1985 and the T-S fuzzymodel-based system subsequently emerged as one of themostactive and fruitful areas of fuzzy control Using a T-S fuzzymodel-based controller a complex dynamic model can bedecomposed into a set of local linear subsystems via fuzzyinference Stability analysis is carried out using the Lyapunovdirect method where the control problem is formulated intolinear matrix inequalities (LMIs) Based on this approachAhn and Anh [9] also developed an inverse double nonlinearautoregressive model with exogenous control based on theT-S fuzzy model applied in a PAM robot A novel 119867
infin
control structure based on a Takagi-Sugeno model [10] wasproposed to track the desired trajectories and simulationresults illustrated the efficiency of the proposed approach forthe new rehabilitation device
The leg rehabilitation machine driven by PAMs is a two-input two-output system This paper proposes compositefuzzy theory which includes T-S fuzzy tracking control andsupervisor control in order to improve tracking performanceThe proposed approach decomposes the model of a nonlin-ear system into a set of linear subsystems with associatednonlinear weighting functions enabling the use of simplelinear control techniques without the need for complicatednonlinear control strategies and also provides a systematicframework for the design of a state feedback controller [11]It has been shown that a composite fuzzy control systemcan be guaranteed to be asymptotically stable if a commonpositive definite solution exists for a set of Lyapunov inequal-ities In addition the supervisory control can overcome thecoupling effect due to two-joint motion In view of the aboveadvantages the proposed controller was applied to the outputtracking control of this system and experimental resultsverified that the proposed controller is capable of achievingexcellent tracking performance
The remainder of the paper is organized as followsSection 2 describes the control strategies Section 3 describesthe system In Section 4 the dynamics of the model arederived Experimental results for output tracking are shownin Section 5 Finally conclusions are presented in Section 6
2 Control Strategies
21 Takagi-Sugeno Fuzzy Tracking Controller Consider ageneral nonlinear dynamic equation
(119905) = 119891 (119909 (119905)) + 119892 (119909 (119905)) 119906 (119905)
119910 (119905) = 119902 (119909 (119905)) (1)
where 119909 isin 119877119899 is the state vector 119910 isin 119877
119898 is the controlledoutput 119906 isin 119877
119898 is the control input vector and119891(119909)119892(119909) and119902(119909) are nonlinear functions with appropriate dimensionsThe nonlinear system (1) can then be expressed by the fuzzysystem
Rule 119894
IF 1199111(119905) is 119865
1119894and sdot sdot sdot and 119911
119892(119905) is 119865
119892119894
THEN (119905) = 119860119894119909 (119905) + 119861
119894119906 (119905) 119894 = 1 2 119903
(2)
where 119911(119905)1
sim 119911119892(119905) are the premise variables including
system states 119865119894119895denotes the fuzzy sets 119903 is the number
of fuzzy rules and 119860119894and 119861
119894are system matrices with
appropriate dimensions For simplicity this study assumedthat the membership functions had been normalized that issum119903
119894=1Π119892
119895=1119865119895119894(119911119895) = 1 As in (1) using the singleton fuzzier
product inferred and weighted defuzzier the fuzzy system isinferred as
(119905) =
119903
sum
119894=1
ℎ119894[119860119894119909 (119905) + 119861
119894119906 (119905)] (3)
where ℎ119894(119911(119905)) = Π
119892
119895=1119865119895119894(119911119895(119905)) Note that sum119903
119894=1ℎ119894(119911(119905)) = 1
for all 119905 wheresum119903119894=1
ℎ119894(119911(119905)) ge 0 for 119894 = 1 2 119903 are regarded
as grade functionsFor output tracking control the control objective is
required to satisfy
119910 (119905) minus 119903 (119905) 997888rarr 0 as 119905 997888rarr infin (4)
where 119903(119905) denotes the desired trajectory or reference signalTo convert the output tracking problem into a stabilizationproblem a set of virtual desired variables 119909
119889(119905) was intro-
duced to be tracked by the state variable 119909 Let 119909(119905) = 119909(119905) minus
119909119889(119905) denote the tracking error for the state variables The
time derivative of 119909(119905) yields
119909 (119905) = minus 119889=
119903
sum
119894
ℎ119894[119860119894119909 (119905) + 119861
119894119906 (119905)] minus
119889(119905) (5)
If the control input 119906(119905) is assumed to satisfy the followingequation
119903
sum
119894=1
ℎ119894119861119894120591 (119905) =
119903
sum
119894=1
ℎ119894119861119894119906 (119905) +
119903
sum
119894=1
ℎ119894119860119894119909119889(119905) minus
119889(119905) (6)
where 120591(119905) is a new control to be designed then the trackingerror system (5) results in the following form
119909 (119905) =
119903
sum
119894=1
ℎ119894119860119894119909 (119905) +
119903
sum
119894=1
ℎ119894119861119894120591 (119905) (7)
The design of the new control 120591(119905) is similar to solvinga stabilization problem The purpose is to steer 119909(119905) to zerowhich means that state 119909(119905) tracks 119909
119889(119905) The new fuzzy
The Scientific World Journal 3
controller 120591(119905) is designed on the basis of parallel distributedcompensation (PDC) and is represented as follows
Rule 119894
IF 1199111(119905) is 119865
1119894and sdot sdot sdot and 119911
119892(119905) is 119865
119892119894
THEN 120591 (119905) = minus119870119894119909 (119905)
(8)
where119870119894represents feedback gainThe inferred output of the
PDC controller is expressed in the following form
120591 (119905) = minus
119903
sum
119894=1
ℎ119894119870119894119909 (119905) (9)
Substituting (9) into (7) yields
119909 (119905) =
119903
sum
119894=1
119903
sum
119895=1
ℎ119894ℎ119895(119860119894minus 119861119894119870119895) 119909 (119905) (10)
The stability analysis of this tracking system (10) is carried outusing the Lyapunov direct method and the Lyapunov func-tion is defined as
119881 (119909 (119905)) = 119909119879
(119905) 119875119909 (119905) gt 0 (11)
where 119875 is a positive symmetric matrix Taking the derivativeof 119881 with respect to time yields
(119909 (119905)) = 119909119879
(119905) 119875119909 + 119909119879
(119905) 119875 119909 (119905)
=
119903
sum
119894=1
119903
sum
119895=1
ℎ119894ℎ119895(119909119879
(119860119879
119894minus 119870119879
119895119861119879
119894) 119875119909)
+
119903
sum
119894=1
119903
sum
119895=1
ℎ119894ℎ119895119909119879
119875 (119860119894minus 119861119894119870119895) 119909
=
119903
sum
119894=1
119903
sum
119895=1
ℎ119894ℎ119895119909119879
[(119860119894minus 119861119894119870119894)119879
119875 + 119875 (119860119894minus 119861119894119870119894)] 119909
(12)
The controller is stable if lt 0 Hence the LMI form isexpressed as follows
(119860119894minus 119861119894119870119894)119879
119875 + 119875 (119860119894minus 119861119894119870119894) lt 0 for 119894 = 1 2 119903
119866119894119895119875 + 119875119866
119894119895lt 0 1 le 119894 lt 119895 le 119903
(13)
where 119866119894119895= (119860119894minus 119861119894119870119895+ 119860119895minus 119861119895119870119894)2 and 119866
119894119894= 119860119894minus 119861119894119870119894
The controller gain 119870119894is obtained using the LMI toolbox
ofMATLAB If there exists a commonpositive definitematrix119875 that satisfies inequalities (13) it can be guaranteed that thetracking error will approach zero
22 Composite Fuzzy Tracking Controller Because the legrehabilitation machine has a coupling effect due to mecha-nism interaction many fuzzymodel controllers in the relatedliterature exhibit restrictive tracking control in application
The proposed approach introduces supervisory control inorder to overcome the coupling effect The 119894th rule of theproposed controller is defined as follows
Rule 119894
IF 1199111(119905) is 119865
1119894and sdot sdot sdot and 119911
119892(119905) is 119865
119892119894
THEN 119906 (119905) = 119870119894119909 (119905) + 119906
119904119894 = 1 2 119903
(14)
where 119906119904(119905) isin 119877
119898 The proposed controller consists of a localstate feedback 119870
119894119909 and a supervisory control 119906
119904 Therefore
the output of the proposed controller is
119906 (119905) =
119903
sum
119894
ℎ119894119870119894119909 (119905) + 119906
119904 (15)
The closed-loop system is given by
(119905) =
119903
sum
119894=1
119903
sum
119895=1
ℎ119894ℎ119895[119860119894minus 119861119894119870119895] 119909 (119905) +
119903
sum
119894=1
ℎ119894119861119894119906119904(119905)
=
119903
sum
119894=1
ℎ2
119894119866119894119894119909 (119905) + 2
119903
sum
119894lt119895
ℎ119894ℎ119895119866119894119895119909 (119905) + 119861119906
119904(119905)
(16)
Suppose that there exist a symmetric and positive definitematrix 119875 and some matrices119870
119894so that the following reduced
stability condition holds
(119860119894minus 119861119894119870119894)119879
119875 + 119875 (119860119894minus 119861119894119870119894) le minus119876
119894 119894 = 1 119903 (17)
where 119876119894is a positive definite matrix Based on this assump-
tion each subsystem is locally controllable and a stablefeedback gain is obtainable Intuitively a common matrix 119875
that satisfies (17) can be obtained more easily than can onethat fulfills the basic stabilization conditions When the LMImethod is applied conditions (17) can be efficiently verified Ifa feasible solution is obtained the design proceeds to exploitthe supervisory control in order to deal with the couplingeffects
Choose the Lyapunov function candidate 1198811(119909) = 119909
119879
119875119909The time derivative of 119881
1(119909) is as follows
1(119909) =
119903
sum
119894=1
ℎ2
119894119909119879
(119866119879
119894119894119875 + 119875119866
119894119894) 119909
+ 2
119903
sum
119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909 + 2119909
119879
119875119861119906119904
le minus
119903
sum
119894=1
ℎ2
119894119909119879
119876119894119909
+ 2
119903
sum
119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909 + 2119909
119879
119875119861119906119904
(18)
Given the matrix property clearly
120582min (119866119879
119894119895119875 + 119875119866
119894119895) 1199092
le 119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909
le 120582min (119866119879
119894119895119875 + 119875119866
119894119895) 1199092
(19)
4 The Scientific World Journal
where 120582min(max) denotes the smallest (largest) eigenvalue ofthe matrix Define
120572 = max119894119895
120582max (119866119879
119894119895119875 + 119875119866
119894119895) for 1 le 119894 lt 119895 le 119903 (20)
A relaxed condition concerning the coupling effect isexpressed as
119903
sum
119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909 le 119896
11199092
1198961=
119903 (119903 minus 1)
2120572
(21)
Finding the maximum value of sum119903119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895)119909
is equivalent to determining the maximum value ofsum119903
119894lt119895ℎ119894ℎ119895120582max(119866
119879
119894119895119875 + 119875119866
119894119895) This can be presented as
a nonlinear programming The optimal algorithms areemployed to seek the best solution Moreover the MATLABOptimization Toolbox consists of functions that minimize ormaximize general nonlinear functions By using the toolboxthe nonlinear programming is expressed in the followingform
max119894119895
119903
sum
119894lt119895
ℎ119894ℎ119895120582max (119866
119879
119894119895119875 + 119875119866
119894119895) 1 le 119894 lt 119895 le 119903
Subject to119903
sum
119894=1
120583119894= 1 120583
119894ge 0
119903
sum
119895=1
120583119895= 1 120583
119895ge 0
(22)
The largest eigenvalue of (119866119879119894119895119875 + 119875119866
119894119895) can be obtained in
advance so the maximum value is determined to be
1198962= max119894119895
sum
119894lt119895
ℎ119894ℎ119895120582max (119866
119879
119894119895119875 + 119875119866
119894119895) (23)
The following supervisory control is chosen
119906119904=
minus119861119879
119875119909
1003817100381710038171003817119909119879119875119861
1003817100381710038171003817
21198961199092
if 10038171003817100381710038171003817119909119879
11987511986110038171003817100381710038171003817
= 0
0 if 10038171003817100381710038171003817119909119879
11987511986110038171003817100381710038171003817= 0
(24)
where 119896 gt 119896119895 119895 = 1 or 2 If 119909119879119875119861 = 0 then substituting (24)
into (18) gives
1(119909) le minus
119903
sum
119894=1
ℎ2
119894119909119879
119876119894119909 + 2119896
1198951199092
minus 21198961199092
le minus
119903
sum
119894=1
ℎ2
119894119909119879
119876119894119909 = minus119881
2(119909)
(25)
where1198812(119909) is a positive definite functionWhen 119909
119879
119875119861 = 0
can give the following form
119909119879
[(119860119894minus 119861119894119870119894)119879
119875 + 119875 (119860119894minus 119861119894119870119894)] 119909
= 119909119879
(119860119879
119894119875 + 119875119860
119894) 119909 le minus119909
119879
119876119894119909 119894 = 1 119903
119903
sum
119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909
=
119903
sum
119894lt119895
ℎ119894ℎ119895[119909119879
(119860119894119875 minus 119875119860
119894) 119909 + 119909
119879
(119860119879
119895119875 minus 119875119860
119895) 119909]
le minussum
119894lt119895
ℎ119894ℎ119895119909119879
(119876119894+ 119876119891) 119909 1 le 119894 lt 119895 le 119903
(26)
the time derivative of 1198811(119909) becomes
1(119909) le minus
119903
sum
119894=1
ℎ2
119894119909119879
119876119894119909 minussum
119894lt119895
ℎ119894ℎ119895119909119879
(119876119894+ 119876119891) 119909 = minus119881
3(119909)
(27)
where 1198813(119909) is a positive definite function Thus the closed-
loop fuzzy system is asymptotically stable
3 System Descriptions
Figure 1 shows the experimental setup including four PAMstwo rotary potentiometers four pressure proportional valvesand four pressure transducers The hardware includes anIBM-compatible personal computer to calculate the con-trol signal which controls the pressure proportional valvethrough a DA card The angles of the joints are detectedusing rotary potentiometers the air pressure of each PAM ismeasured using pressure transducers and the measurementsare then fed back to the computer through anAD cardThesespecifications are listed in Table 1
Figure 2 presents the operation principle of the leg reha-bilitation machine depicting a two-joint leg The behavior ofthe leg manipulated by the rehabilitation machine is similarto that of a human leg Output angles 120579
1and 1205792simulate the
knee and ankle joints and the ranges of the rotary angles 1205791
and 1205792are fromminus45
∘ to 45∘ and fromminus50∘ to 50∘ respectively
The link mass 1198981= 27 kg119898
2= 081 kg and the link length
1198971= 05m 119897
2= 026m The rotating torque 120591 is generated
by the difference in pressure Δ119901 between the two opposingPAMs That is when 119901
119886gt 119901119887 as in Figure 2 the torque
exerted on the joint is counterclockwise and the rotation ofthe joint is also counterclockwise
So a pair of such PAMs is tied together around a pulleywith a radius 119903
119894 as in Figure 2 Then the torque values
imparted to the pulley by the PAM pair are [12]
1205911= (1206011119886
minus 1206011119887) 1199031
1205912= (1206012119886
minus 1206012119887) 1199032
(28)
The Scientific World Journal 5
1
2
3
4
5
6
78910
11
12
1314
15
16
ADDA
Figure 1 The experimental setup
1b
1205911
2b
1205912 m2
r1
1205792
2a
1a
m1
r1
1205791
l1
l2
p1b = p10 minus Δp1
p1a = p10 + Δp1
p2a=p 20
+ Δp2
p 2b=p 20
minus Δp2
Figure 2 Operation principle of the leg rehabilitation machine driven by PAMs
6 The Scientific World Journal
Table 1 Component specifications
Number Component Specifications1 2 3 4 PAM Festo MAS-20-150N7 8 9 10 Pressure proportional valve Mac PPC5C5 6 Rotary potentiometer Keen Engineering KRT205011 12 13 14 Pressure transducer Jihsense SN 911166 0ndash10 kgfcm2
16 IBM-compatible PC Pentium 18GHzDAC ADC Automation AIO3321
where
1206011119886
= 1198651119886(1199011119886) minus 1198701119886(1199011119886) 11990311205791minus 1198611119886(1199011119886) 1199031
1205791
(29a)
1206011119887
= 1198651119887(1199011119887) minus 1198701119887(1199011119887) 11990311205791minus 1198611119887(1199011119887) 1199031
1205791
(29b)
1206012119886
= 1198652119886(1199012119886) minus 1198702119886(1199012119886) 11990321205792minus 1198612119886(1199012119886) 1199032
1205792
(29c)
1206012119887
= 1198652119887(1199012119887) minus 1198702119887(1199012119887) 11990321205792minus 1198612119887(1199012119887) 1199032
1205792 (29d)
where the spring coefficient119870(119901) and the damping coefficient119861(119901) are given by Reynolds et al [13]
The desired input pressures P119886= [1199011119886
1199012119886]119879 and P
119887=
[1199011119887
1199012119887]119879 for each PAM are generated by the following
equation
P119886(119905) = P
0+ ΔP (119905) P
119887(119905) = P
0minus ΔP (119905) (30)
where P0
= [11990110
11990120]119879 is a nominal constant input PAM
pressure and ΔP(119905) = [Δ1199011
Δ1199012]119879 is the control pressure
input with an arbitrary function of time Because the pressureinputΔP(119905) = [Δ119901
1Δ1199012]119879 and output 120579 = [120579
11205792]119879 the sys-
tem can bewritten as a two-input two-output (TITO) controlsystem The control signal u = [119906
11199062]119879 is proportional to
ΔP based on the pressure proportional valversquos characteristicsThat is ΔP can be used instead of u as a control input
4 Dynamic Model of a Two-Joint LegRehabilitation Machine Driven by PAMs
Figure 2 shows a two-joint leg rehabilitation machine drivenby PAMs and the dynamic equation is given as follows [14]
119872(120579) 120579 + 119862 (120579 120579) 120579 + 119866 (120579) = 120591 (31)
where
119872(120579) = [(1198981+ 1198982) 1198972
119898211989711198972(11990411199042+ 11988811198882)
119898211989711198972(11990411199042+ 11988811198882) 119898
21198972
2
]
119862 (120579 120579) = [0 minus119898
211989711198972(11988811199042minus 11990411198882) 1205792
minus119898211989711198972(11988811199042minus 11990411198882) 1205791
0]
119866 (120579) = [(1198981+ 1198982) 11989711198921199041
minus119898211989721198921199042
]
(32)
and119872(120579) is the moment of inertia 119862(120579 120579) includes Coriolisand centripetal force and 119866(120579) is the gravitational force
Notation 1199041= sin(120579
1) 1199042= sin(120579
2) 1198881= cos(120579
1) and 119888
2=
cos(1205792) Let 119909
1= 1205791 1199092= 1205791 1199093= 1205792 and 119909
4= 1205792 then (31)
can be written as the following state-space form [14]
1= 1199092
2= 1198911(119909) + 119892
11(119909) 1205911+ 119892121205912
3= 1199094
4= 1198912(119909) + 119892
21(119909) + 119892
221205912
(33)
where1198911(119909)
=
(11990411198882minus 11988811199042) [119898211989711198972(11990411199042+ 11988811198882) 1199092
2minus 11989821198972
21199092
4]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
+[(1198981minus 1198982) 11989721198921199041minus 119898211989721198921199042(11990411199042+ 11988811198882)]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
1198912(119909)
=
(11990411198882minus 11988811199042) [minus (119898
1+ 1198982) 1198972
11199092
2+ 119898211989711198972(11990411199042+ 11988811198882) 1199092
4]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
+[minus (119898
1+ 1198982) 11989711198921199041(11990411199042+ 11988811198882) + (119898
1+ 1198982) 11989711198921199042]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989211(119909) =
11989821198972
2
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989212(119909) =
minus119898211989711198972(11990411199042+ 11988811198882)
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989221(119909) =
minus119898211989711198972(11990411199042+ 11988811198882)
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989222(119909) =
(1198981+ 1198982) 1198972
1
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
(34)
5 Experimental Studies
Figure 3 shows the leg rehabilitation machine using an actualhuman loading with a 65 kg weight The automatic device
The Scientific World Journal 7
Figure 3 The two-joint leg rehabilitation device with actual human loading
can help patients to recover lower limb motion function bymeans of continuous passive motion such as a sinusoidalwave command an irregular curve command and an end-effect tracking command The experiments include both theproposed approach and PDC for comparison in order toevaluate efficacy and control performance The controllerswere implemented on an Intel Pentium 18GHz PC with asampling time of 5ms and the entire control software wascoded in C++
This study attempts to use as few rules as possible inorder tominimize design effort and complexityTheT-S fuzzymodel of the system is thus given the following four-rulefuzzy model
1198771
IF 1199091is about (1205874) and 119909
3is about (1205874)
THEN = 11986011199091+ 1198611119906
1198772
IF 1199091is about (1205874) and 119909
3is about (minus1205874)
THEN = 11986021199091+ 1198612119906
1198773
IF 1199091is about (minus1205874) and 119909
3is about (1205874)
THEN = 11986031199091+ 1198613119906
1198774
IF 1199091is about (minus1205874) and 119909
3is about (minus1205874)
THEN = 11986041199091+ 1198614119906
(35)
where
1198601=
[[[
[
0 1 0 0
44082 00059 06742 minus00002
0 0 0 1
minus14572 00002 minus39722 00002
]]]
]
1198602=
[[[
[
0 1 0 0
46734 00049 0532 00001
0 0 0 1
minus12145 minus00001 minus3563 00001
]]]
]
1198603=
[[[
[
0 1 0 0
54782 00021 06876 minus00002
0 0 0 1
14525 00002 46723 00002
]]]
]
1198604=
[[[[
[
0 1 0 0
49821 minus00044 06543 00002
0 0 0 1
minus11034 minus00002 35631 00003
]]]]
]
1198611= 1198614=
[[[[
[
0 0
14811 minus2849
0 0
minus2849 237417
]]]]
]
1198612= 1198613=
[[[[
[
0 0
11965 10297
0 0
10297 191501
]]]]
]
119875 =
[[[[
[
162602 minus06640 10734 minus00095
minus06640 03007 minus03455 02871
10734 minus03455 04730 minus03458
minus00095 02871 minus03458 03658
]]]]
]
1198701= [
minus04509 01745 minus07182 00527
minus02807 00034 minus00471 02639]
1198702= [
minus04619 01238 minus07812 00351
minus02827 00068 minus00851 01401]
1198703= [
minus05902 02132 minus08910 00531
minus04107 00117 minus00730 02989]
1198704= [
minus04891 02109 minus06734 004145
minus03180 00085 minus00631 03893]
(36)
which guarantee the stability condition (17) MATLAB Tool-box is used to obtain parameters as 119896
1= 00251 and
8 The Scientific World Journal
Table 2 Peak-peak error and phase lag for Figure 5
The proposed approach PDCPeak-peak error Phase lag Peak-peak error Phase lag
1205791
1205792
1205791
1205792
1205791
1205792
1205791
1205792
07 035 49∘ 45∘ 13 065 162∘ 108∘
20
10
0
minus10
minus20
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
ReferenceActual
15
0
minus15
1205792
(a) The proposed approach
20
10
0
minus10
minus20
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
ReferenceActual
15
0
minus15
1205792
(b) PDC
Figure 4 Sinusoidal wave response for knee and ankle joints
1198962= 001539 For comparison with the proposed controller
the PDC feedback gains are designed to be
1= [
minus03293 01023 minus02538 002317
minus01783 00021 minus002672 04732]
1198702= [
minus04278 01845 minus02451 00731
minus02185 00026 minus00752 00923]
1198703= [
minus06013 03461 minus05482 00231
minus04421 00093 minus00651 03529]
1198704= [
minus03756 03150 minus05391 00421
minus04250 00023 minus00531 03597]
(37)
51 Sinusoidal Wave Response Continuous reciprocation isrequired in order to foster the recovery of extremity functionThe sinusoidal wave responses of the proposed approach andPDC for both knee and ankle joints are shown in Figure 4It is evident that angle trajectories of the proposed approachare close to the command Figure 5 shows that the proposedapproach exhibits less tracking errors than does PDC Thepeak-peak error and phase lag are listed in Table 2 Because
of the interaction of the two joints PDC has significant angleerrors for 120579
1 which will degrade the rehabilitation effect
However supervisory control can overcome the couplingeffect of the two joints to achieve excellent rehabilitationfunction for patients
52 Irregular Curve Response In practical applications itcould be expected that the reference command will changewith different input frequencies The desired trajectories forboth knee and ankle joints are
1205791= 20 lowast 033 (sin (2120587119891
1119905) + sin (2120587119891
2119905) + sin (2120587119891
3119905))
1205792= 15 lowast 033 (sin (2120587119891
1119905) + sin (2120587119891
2119905) + sin (2120587119891
3119905))
(38)
with 1198911= 005Hz 119891
2= 01Hz and 119891
3= 0066Hz
Figure 6 shows the tracking responses of irregular curvesobtained using both the proposed approach and PDC Track-ing errors for the knee and ankle joints are shown in Figure 7Clearly the angle error of the proposed approach is averagemaintained within 2∘ However the proposed approach iscapable of adapting to different frequencies
The Scientific World Journal 9
14
12
10
8
6
4
2
0
minus2
minus41009080706050403020100
Time (s)
Ang
le (d
eg)
1205791 error1205792 error
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 5 Angle tracking errors for both knee and ankle
20
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
120579215
10
5
0
minus5
minus10
minus15
Actual
(a) The proposed approach
20
10
0
minus10
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
1205792
Actual
(b) PDC
Figure 6 Irregular curve response for both knee and ankle joints
53 Elliptic Response The desired end-effect or trajectory isgiven by
119909119889(119905) = 0614 minus 0015 sdot cos (02120587 sdot 119905 minus 120587)
119910119889(119905) = minus01 sdot sin (02120587 sdot 119905 minus 2120587)
(39)
where 0 le 119905 le 20 seconds
The end-effect tracking responses in the 119909 119910 coordinatefor both the proposed approach and PDC are shown inFigure 8 and the end-effect position tracking errors aredisplayed in Figure 9 It is evident that tracking behavior ofthe proposed approach is better than that of the PDC Ascan be seen the tracking errors of the proposed approach are
10 The Scientific World Journal
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 7 Angle tracking errors for both the proposed approach and PDC
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(a) The proposed approach
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(b) PDC
Figure 8 Elliptic response of the proposed approach and PDC
within 003m On the other hand angle tracking errors ofknee and ankle joints are shown in Figure 10
Moreover it is difficult to enhance end-effector trackingperformance using the PDC algorithm because the PDCcannot overcome the nonlinearity of PAMs and the structuralinteraction However the proposed approach overcomessuccessfully the coupling effect and parameter uncertaintiesof the system As seen in the experimental results theproposed approach can attain excellent end-effector trackingperformance in rehabilitation function
6 Conclusions
In this study a novel composite fuzzy control is proposed andapplied in the two-joint leg rehabilitation device driven byPAMsThe proposed controller is not only capable of decom-posing nonlinear systems into a set of linear subsystems but isalso capable of simplifying a complex nonlinear system usinglinear control techniques with the control gains determinedusingMATLABrsquos LMI Toolbox based on the Lyapunov stabil-ity theoremMoreover the supervisory control can overcome
The Scientific World Journal 11
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(a) The proposed approach
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(b) PDC
Figure 9 The end-effect position tracking errors
14
12
10
8
6
4
2
0
minus2
minus40 2 4 6 8 10 12 14 16 18 20
Ang
le (d
eg)
Time (s)
1205791 error1205792 error
(a) The proposed approach
0 2 4 6 8 10 12 14 16 18 20
Time (s)
15
10
5
0
minus5
minus10
1205791 error1205792 error
Ang
le (d
eg)
(b) PDC
Figure 10 Angle tracking errors of knee and ankle joints
the coupling effect for a leg rehabilitation machine Experi-mental results show that the system response of the proposedapproach was in good agreement with that of the refer-ence input and guarantee robustness to system parameteruncertainties
Conflict of Interests
The author declares that there is no conflict of interestsregarding the publication of this paper
References
[1] D G Caldwell G A Medrano-Cerda and M GoodwinldquoControl of pneumatic muscle actuatorsrdquo IEEE Control SystemsMagazine vol 15 no 1 pp 40ndash48 1995
[2] C P Chou and B Hannaford ldquoStatic and dynamic characteris-tics ofMcKibben pneumatic artificialmusclesrdquo inProceedings ofthe IEEE International Conference on Robotics and Automationpp 281ndash286 May 1994
[3] T Noritsugu and T Tanaka ldquoApplication of rubber artificialmuscle manipulator as a rehabilitation robotrdquo IEEEASMETransactions on Mechatronics vol 2 no 4 pp 259ndash267 1997
12 The Scientific World Journal
[4] J H Lilly and L Yang ldquoSliding mode tracking for pneumaticmuscle actuators in opposing pair configurationrdquo IEEE Trans-actions on Control Systems Technology vol 4 pp 550ndash558 2005
[5] K K Ahn and H P H Anh ldquoDesign and implementation of anadaptive recurrent neural networks (ARNN) controller of thepneumatic artificial muscle (PAM) manipulatorrdquoMechatronicsvol 19 no 6 pp 816ndash828 2009
[6] X Shen ldquoNonlinearmodel-based control of pneumatic artificialmuscle servo systemsrdquo Control Engineering Practice vol 18 no3 pp 311ndash317 2010
[7] L A Zadeh ldquoFuzzy setsrdquo Information and Control vol 8 no 3pp 338ndash353 1965
[8] T Takagi and M Sugeno ldquoFuzzy identification of systems andits applications to modeling and controlrdquo IEEE Transactions onSystems Man and Cybernetics vol 15 no 1 pp 116ndash132 1985
[9] K K Ahn and H P H Anh ldquoInverse double NARX fuzzymodeling for system identificationrdquo IEEEASME Transactionson Mechatronics vol 15 no 1 pp 136ndash148 2010
[10] L Seddiki K Guelton and J Zaytoon ldquoConcept and Takagi-Sugeno descriptor tracking controller design of a closedmuscu-lar chain lower-limb rehabilitation devicerdquo IET Control Theoryand Applications vol 4 no 8 pp 1407ndash1420 2010
[11] K Tanaka T Ikeda and H O Wang ldquoFuzzy regulators andfuzzy observers relaxed stability conditions and LMI-baseddesignsrdquo IEEE Transactions on Fuzzy Systems vol 6 no 2 pp250ndash265 1998
[12] M K Chang J J Liou andM L Chen ldquoT-S fuzzymodel-basedtracking control of a one-dimensional manipulator actuated bypneumatic artificial musclesrdquo Control Engineering Practice vol19 no 12 pp 1442ndash1449 2011
[13] D B Reynolds D W Repperger C A Phillips and G BandryldquoModeling the dynamic characteristics of pneumatic musclerdquoAnnals of Biomedical Engineering vol 31 no 3 pp 310ndash3172003
[14] C S Tseng B S Chen and H J Uang ldquoFuzzy tracking controldesign for nonlinear dynamic systems via T-S fuzzy modelrdquoIEEE Transactions on Fuzzy Systems vol 9 no 3 pp 381ndash3922001
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
2 The Scientific World Journal
were consistent with theoretical findings for two differentmasses Ahn and Anh [5] adopted an ARNN controller ina PAM manipulator for reducing tracking errors Shen [6]developed a full nonlinear model that encompassed allthe major existing nonlinearities Based on this model thestandard slidingmode control approachwas applied to obtainrobust control even in the event of model uncertainties anddisturbances
Since the inception of fuzzy set theory by Zadeh [7] in1965 a great deal of research has been focused on fuzzycontrol systems Takagi and Sugeno [8] proposed the T-S fuzzy model-based controller in 1985 and the T-S fuzzymodel-based system subsequently emerged as one of themostactive and fruitful areas of fuzzy control Using a T-S fuzzymodel-based controller a complex dynamic model can bedecomposed into a set of local linear subsystems via fuzzyinference Stability analysis is carried out using the Lyapunovdirect method where the control problem is formulated intolinear matrix inequalities (LMIs) Based on this approachAhn and Anh [9] also developed an inverse double nonlinearautoregressive model with exogenous control based on theT-S fuzzy model applied in a PAM robot A novel 119867
infin
control structure based on a Takagi-Sugeno model [10] wasproposed to track the desired trajectories and simulationresults illustrated the efficiency of the proposed approach forthe new rehabilitation device
The leg rehabilitation machine driven by PAMs is a two-input two-output system This paper proposes compositefuzzy theory which includes T-S fuzzy tracking control andsupervisor control in order to improve tracking performanceThe proposed approach decomposes the model of a nonlin-ear system into a set of linear subsystems with associatednonlinear weighting functions enabling the use of simplelinear control techniques without the need for complicatednonlinear control strategies and also provides a systematicframework for the design of a state feedback controller [11]It has been shown that a composite fuzzy control systemcan be guaranteed to be asymptotically stable if a commonpositive definite solution exists for a set of Lyapunov inequal-ities In addition the supervisory control can overcome thecoupling effect due to two-joint motion In view of the aboveadvantages the proposed controller was applied to the outputtracking control of this system and experimental resultsverified that the proposed controller is capable of achievingexcellent tracking performance
The remainder of the paper is organized as followsSection 2 describes the control strategies Section 3 describesthe system In Section 4 the dynamics of the model arederived Experimental results for output tracking are shownin Section 5 Finally conclusions are presented in Section 6
2 Control Strategies
21 Takagi-Sugeno Fuzzy Tracking Controller Consider ageneral nonlinear dynamic equation
(119905) = 119891 (119909 (119905)) + 119892 (119909 (119905)) 119906 (119905)
119910 (119905) = 119902 (119909 (119905)) (1)
where 119909 isin 119877119899 is the state vector 119910 isin 119877
119898 is the controlledoutput 119906 isin 119877
119898 is the control input vector and119891(119909)119892(119909) and119902(119909) are nonlinear functions with appropriate dimensionsThe nonlinear system (1) can then be expressed by the fuzzysystem
Rule 119894
IF 1199111(119905) is 119865
1119894and sdot sdot sdot and 119911
119892(119905) is 119865
119892119894
THEN (119905) = 119860119894119909 (119905) + 119861
119894119906 (119905) 119894 = 1 2 119903
(2)
where 119911(119905)1
sim 119911119892(119905) are the premise variables including
system states 119865119894119895denotes the fuzzy sets 119903 is the number
of fuzzy rules and 119860119894and 119861
119894are system matrices with
appropriate dimensions For simplicity this study assumedthat the membership functions had been normalized that issum119903
119894=1Π119892
119895=1119865119895119894(119911119895) = 1 As in (1) using the singleton fuzzier
product inferred and weighted defuzzier the fuzzy system isinferred as
(119905) =
119903
sum
119894=1
ℎ119894[119860119894119909 (119905) + 119861
119894119906 (119905)] (3)
where ℎ119894(119911(119905)) = Π
119892
119895=1119865119895119894(119911119895(119905)) Note that sum119903
119894=1ℎ119894(119911(119905)) = 1
for all 119905 wheresum119903119894=1
ℎ119894(119911(119905)) ge 0 for 119894 = 1 2 119903 are regarded
as grade functionsFor output tracking control the control objective is
required to satisfy
119910 (119905) minus 119903 (119905) 997888rarr 0 as 119905 997888rarr infin (4)
where 119903(119905) denotes the desired trajectory or reference signalTo convert the output tracking problem into a stabilizationproblem a set of virtual desired variables 119909
119889(119905) was intro-
duced to be tracked by the state variable 119909 Let 119909(119905) = 119909(119905) minus
119909119889(119905) denote the tracking error for the state variables The
time derivative of 119909(119905) yields
119909 (119905) = minus 119889=
119903
sum
119894
ℎ119894[119860119894119909 (119905) + 119861
119894119906 (119905)] minus
119889(119905) (5)
If the control input 119906(119905) is assumed to satisfy the followingequation
119903
sum
119894=1
ℎ119894119861119894120591 (119905) =
119903
sum
119894=1
ℎ119894119861119894119906 (119905) +
119903
sum
119894=1
ℎ119894119860119894119909119889(119905) minus
119889(119905) (6)
where 120591(119905) is a new control to be designed then the trackingerror system (5) results in the following form
119909 (119905) =
119903
sum
119894=1
ℎ119894119860119894119909 (119905) +
119903
sum
119894=1
ℎ119894119861119894120591 (119905) (7)
The design of the new control 120591(119905) is similar to solvinga stabilization problem The purpose is to steer 119909(119905) to zerowhich means that state 119909(119905) tracks 119909
119889(119905) The new fuzzy
The Scientific World Journal 3
controller 120591(119905) is designed on the basis of parallel distributedcompensation (PDC) and is represented as follows
Rule 119894
IF 1199111(119905) is 119865
1119894and sdot sdot sdot and 119911
119892(119905) is 119865
119892119894
THEN 120591 (119905) = minus119870119894119909 (119905)
(8)
where119870119894represents feedback gainThe inferred output of the
PDC controller is expressed in the following form
120591 (119905) = minus
119903
sum
119894=1
ℎ119894119870119894119909 (119905) (9)
Substituting (9) into (7) yields
119909 (119905) =
119903
sum
119894=1
119903
sum
119895=1
ℎ119894ℎ119895(119860119894minus 119861119894119870119895) 119909 (119905) (10)
The stability analysis of this tracking system (10) is carried outusing the Lyapunov direct method and the Lyapunov func-tion is defined as
119881 (119909 (119905)) = 119909119879
(119905) 119875119909 (119905) gt 0 (11)
where 119875 is a positive symmetric matrix Taking the derivativeof 119881 with respect to time yields
(119909 (119905)) = 119909119879
(119905) 119875119909 + 119909119879
(119905) 119875 119909 (119905)
=
119903
sum
119894=1
119903
sum
119895=1
ℎ119894ℎ119895(119909119879
(119860119879
119894minus 119870119879
119895119861119879
119894) 119875119909)
+
119903
sum
119894=1
119903
sum
119895=1
ℎ119894ℎ119895119909119879
119875 (119860119894minus 119861119894119870119895) 119909
=
119903
sum
119894=1
119903
sum
119895=1
ℎ119894ℎ119895119909119879
[(119860119894minus 119861119894119870119894)119879
119875 + 119875 (119860119894minus 119861119894119870119894)] 119909
(12)
The controller is stable if lt 0 Hence the LMI form isexpressed as follows
(119860119894minus 119861119894119870119894)119879
119875 + 119875 (119860119894minus 119861119894119870119894) lt 0 for 119894 = 1 2 119903
119866119894119895119875 + 119875119866
119894119895lt 0 1 le 119894 lt 119895 le 119903
(13)
where 119866119894119895= (119860119894minus 119861119894119870119895+ 119860119895minus 119861119895119870119894)2 and 119866
119894119894= 119860119894minus 119861119894119870119894
The controller gain 119870119894is obtained using the LMI toolbox
ofMATLAB If there exists a commonpositive definitematrix119875 that satisfies inequalities (13) it can be guaranteed that thetracking error will approach zero
22 Composite Fuzzy Tracking Controller Because the legrehabilitation machine has a coupling effect due to mecha-nism interaction many fuzzymodel controllers in the relatedliterature exhibit restrictive tracking control in application
The proposed approach introduces supervisory control inorder to overcome the coupling effect The 119894th rule of theproposed controller is defined as follows
Rule 119894
IF 1199111(119905) is 119865
1119894and sdot sdot sdot and 119911
119892(119905) is 119865
119892119894
THEN 119906 (119905) = 119870119894119909 (119905) + 119906
119904119894 = 1 2 119903
(14)
where 119906119904(119905) isin 119877
119898 The proposed controller consists of a localstate feedback 119870
119894119909 and a supervisory control 119906
119904 Therefore
the output of the proposed controller is
119906 (119905) =
119903
sum
119894
ℎ119894119870119894119909 (119905) + 119906
119904 (15)
The closed-loop system is given by
(119905) =
119903
sum
119894=1
119903
sum
119895=1
ℎ119894ℎ119895[119860119894minus 119861119894119870119895] 119909 (119905) +
119903
sum
119894=1
ℎ119894119861119894119906119904(119905)
=
119903
sum
119894=1
ℎ2
119894119866119894119894119909 (119905) + 2
119903
sum
119894lt119895
ℎ119894ℎ119895119866119894119895119909 (119905) + 119861119906
119904(119905)
(16)
Suppose that there exist a symmetric and positive definitematrix 119875 and some matrices119870
119894so that the following reduced
stability condition holds
(119860119894minus 119861119894119870119894)119879
119875 + 119875 (119860119894minus 119861119894119870119894) le minus119876
119894 119894 = 1 119903 (17)
where 119876119894is a positive definite matrix Based on this assump-
tion each subsystem is locally controllable and a stablefeedback gain is obtainable Intuitively a common matrix 119875
that satisfies (17) can be obtained more easily than can onethat fulfills the basic stabilization conditions When the LMImethod is applied conditions (17) can be efficiently verified Ifa feasible solution is obtained the design proceeds to exploitthe supervisory control in order to deal with the couplingeffects
Choose the Lyapunov function candidate 1198811(119909) = 119909
119879
119875119909The time derivative of 119881
1(119909) is as follows
1(119909) =
119903
sum
119894=1
ℎ2
119894119909119879
(119866119879
119894119894119875 + 119875119866
119894119894) 119909
+ 2
119903
sum
119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909 + 2119909
119879
119875119861119906119904
le minus
119903
sum
119894=1
ℎ2
119894119909119879
119876119894119909
+ 2
119903
sum
119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909 + 2119909
119879
119875119861119906119904
(18)
Given the matrix property clearly
120582min (119866119879
119894119895119875 + 119875119866
119894119895) 1199092
le 119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909
le 120582min (119866119879
119894119895119875 + 119875119866
119894119895) 1199092
(19)
4 The Scientific World Journal
where 120582min(max) denotes the smallest (largest) eigenvalue ofthe matrix Define
120572 = max119894119895
120582max (119866119879
119894119895119875 + 119875119866
119894119895) for 1 le 119894 lt 119895 le 119903 (20)
A relaxed condition concerning the coupling effect isexpressed as
119903
sum
119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909 le 119896
11199092
1198961=
119903 (119903 minus 1)
2120572
(21)
Finding the maximum value of sum119903119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895)119909
is equivalent to determining the maximum value ofsum119903
119894lt119895ℎ119894ℎ119895120582max(119866
119879
119894119895119875 + 119875119866
119894119895) This can be presented as
a nonlinear programming The optimal algorithms areemployed to seek the best solution Moreover the MATLABOptimization Toolbox consists of functions that minimize ormaximize general nonlinear functions By using the toolboxthe nonlinear programming is expressed in the followingform
max119894119895
119903
sum
119894lt119895
ℎ119894ℎ119895120582max (119866
119879
119894119895119875 + 119875119866
119894119895) 1 le 119894 lt 119895 le 119903
Subject to119903
sum
119894=1
120583119894= 1 120583
119894ge 0
119903
sum
119895=1
120583119895= 1 120583
119895ge 0
(22)
The largest eigenvalue of (119866119879119894119895119875 + 119875119866
119894119895) can be obtained in
advance so the maximum value is determined to be
1198962= max119894119895
sum
119894lt119895
ℎ119894ℎ119895120582max (119866
119879
119894119895119875 + 119875119866
119894119895) (23)
The following supervisory control is chosen
119906119904=
minus119861119879
119875119909
1003817100381710038171003817119909119879119875119861
1003817100381710038171003817
21198961199092
if 10038171003817100381710038171003817119909119879
11987511986110038171003817100381710038171003817
= 0
0 if 10038171003817100381710038171003817119909119879
11987511986110038171003817100381710038171003817= 0
(24)
where 119896 gt 119896119895 119895 = 1 or 2 If 119909119879119875119861 = 0 then substituting (24)
into (18) gives
1(119909) le minus
119903
sum
119894=1
ℎ2
119894119909119879
119876119894119909 + 2119896
1198951199092
minus 21198961199092
le minus
119903
sum
119894=1
ℎ2
119894119909119879
119876119894119909 = minus119881
2(119909)
(25)
where1198812(119909) is a positive definite functionWhen 119909
119879
119875119861 = 0
can give the following form
119909119879
[(119860119894minus 119861119894119870119894)119879
119875 + 119875 (119860119894minus 119861119894119870119894)] 119909
= 119909119879
(119860119879
119894119875 + 119875119860
119894) 119909 le minus119909
119879
119876119894119909 119894 = 1 119903
119903
sum
119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909
=
119903
sum
119894lt119895
ℎ119894ℎ119895[119909119879
(119860119894119875 minus 119875119860
119894) 119909 + 119909
119879
(119860119879
119895119875 minus 119875119860
119895) 119909]
le minussum
119894lt119895
ℎ119894ℎ119895119909119879
(119876119894+ 119876119891) 119909 1 le 119894 lt 119895 le 119903
(26)
the time derivative of 1198811(119909) becomes
1(119909) le minus
119903
sum
119894=1
ℎ2
119894119909119879
119876119894119909 minussum
119894lt119895
ℎ119894ℎ119895119909119879
(119876119894+ 119876119891) 119909 = minus119881
3(119909)
(27)
where 1198813(119909) is a positive definite function Thus the closed-
loop fuzzy system is asymptotically stable
3 System Descriptions
Figure 1 shows the experimental setup including four PAMstwo rotary potentiometers four pressure proportional valvesand four pressure transducers The hardware includes anIBM-compatible personal computer to calculate the con-trol signal which controls the pressure proportional valvethrough a DA card The angles of the joints are detectedusing rotary potentiometers the air pressure of each PAM ismeasured using pressure transducers and the measurementsare then fed back to the computer through anAD cardThesespecifications are listed in Table 1
Figure 2 presents the operation principle of the leg reha-bilitation machine depicting a two-joint leg The behavior ofthe leg manipulated by the rehabilitation machine is similarto that of a human leg Output angles 120579
1and 1205792simulate the
knee and ankle joints and the ranges of the rotary angles 1205791
and 1205792are fromminus45
∘ to 45∘ and fromminus50∘ to 50∘ respectively
The link mass 1198981= 27 kg119898
2= 081 kg and the link length
1198971= 05m 119897
2= 026m The rotating torque 120591 is generated
by the difference in pressure Δ119901 between the two opposingPAMs That is when 119901
119886gt 119901119887 as in Figure 2 the torque
exerted on the joint is counterclockwise and the rotation ofthe joint is also counterclockwise
So a pair of such PAMs is tied together around a pulleywith a radius 119903
119894 as in Figure 2 Then the torque values
imparted to the pulley by the PAM pair are [12]
1205911= (1206011119886
minus 1206011119887) 1199031
1205912= (1206012119886
minus 1206012119887) 1199032
(28)
The Scientific World Journal 5
1
2
3
4
5
6
78910
11
12
1314
15
16
ADDA
Figure 1 The experimental setup
1b
1205911
2b
1205912 m2
r1
1205792
2a
1a
m1
r1
1205791
l1
l2
p1b = p10 minus Δp1
p1a = p10 + Δp1
p2a=p 20
+ Δp2
p 2b=p 20
minus Δp2
Figure 2 Operation principle of the leg rehabilitation machine driven by PAMs
6 The Scientific World Journal
Table 1 Component specifications
Number Component Specifications1 2 3 4 PAM Festo MAS-20-150N7 8 9 10 Pressure proportional valve Mac PPC5C5 6 Rotary potentiometer Keen Engineering KRT205011 12 13 14 Pressure transducer Jihsense SN 911166 0ndash10 kgfcm2
16 IBM-compatible PC Pentium 18GHzDAC ADC Automation AIO3321
where
1206011119886
= 1198651119886(1199011119886) minus 1198701119886(1199011119886) 11990311205791minus 1198611119886(1199011119886) 1199031
1205791
(29a)
1206011119887
= 1198651119887(1199011119887) minus 1198701119887(1199011119887) 11990311205791minus 1198611119887(1199011119887) 1199031
1205791
(29b)
1206012119886
= 1198652119886(1199012119886) minus 1198702119886(1199012119886) 11990321205792minus 1198612119886(1199012119886) 1199032
1205792
(29c)
1206012119887
= 1198652119887(1199012119887) minus 1198702119887(1199012119887) 11990321205792minus 1198612119887(1199012119887) 1199032
1205792 (29d)
where the spring coefficient119870(119901) and the damping coefficient119861(119901) are given by Reynolds et al [13]
The desired input pressures P119886= [1199011119886
1199012119886]119879 and P
119887=
[1199011119887
1199012119887]119879 for each PAM are generated by the following
equation
P119886(119905) = P
0+ ΔP (119905) P
119887(119905) = P
0minus ΔP (119905) (30)
where P0
= [11990110
11990120]119879 is a nominal constant input PAM
pressure and ΔP(119905) = [Δ1199011
Δ1199012]119879 is the control pressure
input with an arbitrary function of time Because the pressureinputΔP(119905) = [Δ119901
1Δ1199012]119879 and output 120579 = [120579
11205792]119879 the sys-
tem can bewritten as a two-input two-output (TITO) controlsystem The control signal u = [119906
11199062]119879 is proportional to
ΔP based on the pressure proportional valversquos characteristicsThat is ΔP can be used instead of u as a control input
4 Dynamic Model of a Two-Joint LegRehabilitation Machine Driven by PAMs
Figure 2 shows a two-joint leg rehabilitation machine drivenby PAMs and the dynamic equation is given as follows [14]
119872(120579) 120579 + 119862 (120579 120579) 120579 + 119866 (120579) = 120591 (31)
where
119872(120579) = [(1198981+ 1198982) 1198972
119898211989711198972(11990411199042+ 11988811198882)
119898211989711198972(11990411199042+ 11988811198882) 119898
21198972
2
]
119862 (120579 120579) = [0 minus119898
211989711198972(11988811199042minus 11990411198882) 1205792
minus119898211989711198972(11988811199042minus 11990411198882) 1205791
0]
119866 (120579) = [(1198981+ 1198982) 11989711198921199041
minus119898211989721198921199042
]
(32)
and119872(120579) is the moment of inertia 119862(120579 120579) includes Coriolisand centripetal force and 119866(120579) is the gravitational force
Notation 1199041= sin(120579
1) 1199042= sin(120579
2) 1198881= cos(120579
1) and 119888
2=
cos(1205792) Let 119909
1= 1205791 1199092= 1205791 1199093= 1205792 and 119909
4= 1205792 then (31)
can be written as the following state-space form [14]
1= 1199092
2= 1198911(119909) + 119892
11(119909) 1205911+ 119892121205912
3= 1199094
4= 1198912(119909) + 119892
21(119909) + 119892
221205912
(33)
where1198911(119909)
=
(11990411198882minus 11988811199042) [119898211989711198972(11990411199042+ 11988811198882) 1199092
2minus 11989821198972
21199092
4]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
+[(1198981minus 1198982) 11989721198921199041minus 119898211989721198921199042(11990411199042+ 11988811198882)]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
1198912(119909)
=
(11990411198882minus 11988811199042) [minus (119898
1+ 1198982) 1198972
11199092
2+ 119898211989711198972(11990411199042+ 11988811198882) 1199092
4]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
+[minus (119898
1+ 1198982) 11989711198921199041(11990411199042+ 11988811198882) + (119898
1+ 1198982) 11989711198921199042]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989211(119909) =
11989821198972
2
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989212(119909) =
minus119898211989711198972(11990411199042+ 11988811198882)
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989221(119909) =
minus119898211989711198972(11990411199042+ 11988811198882)
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989222(119909) =
(1198981+ 1198982) 1198972
1
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
(34)
5 Experimental Studies
Figure 3 shows the leg rehabilitation machine using an actualhuman loading with a 65 kg weight The automatic device
The Scientific World Journal 7
Figure 3 The two-joint leg rehabilitation device with actual human loading
can help patients to recover lower limb motion function bymeans of continuous passive motion such as a sinusoidalwave command an irregular curve command and an end-effect tracking command The experiments include both theproposed approach and PDC for comparison in order toevaluate efficacy and control performance The controllerswere implemented on an Intel Pentium 18GHz PC with asampling time of 5ms and the entire control software wascoded in C++
This study attempts to use as few rules as possible inorder tominimize design effort and complexityTheT-S fuzzymodel of the system is thus given the following four-rulefuzzy model
1198771
IF 1199091is about (1205874) and 119909
3is about (1205874)
THEN = 11986011199091+ 1198611119906
1198772
IF 1199091is about (1205874) and 119909
3is about (minus1205874)
THEN = 11986021199091+ 1198612119906
1198773
IF 1199091is about (minus1205874) and 119909
3is about (1205874)
THEN = 11986031199091+ 1198613119906
1198774
IF 1199091is about (minus1205874) and 119909
3is about (minus1205874)
THEN = 11986041199091+ 1198614119906
(35)
where
1198601=
[[[
[
0 1 0 0
44082 00059 06742 minus00002
0 0 0 1
minus14572 00002 minus39722 00002
]]]
]
1198602=
[[[
[
0 1 0 0
46734 00049 0532 00001
0 0 0 1
minus12145 minus00001 minus3563 00001
]]]
]
1198603=
[[[
[
0 1 0 0
54782 00021 06876 minus00002
0 0 0 1
14525 00002 46723 00002
]]]
]
1198604=
[[[[
[
0 1 0 0
49821 minus00044 06543 00002
0 0 0 1
minus11034 minus00002 35631 00003
]]]]
]
1198611= 1198614=
[[[[
[
0 0
14811 minus2849
0 0
minus2849 237417
]]]]
]
1198612= 1198613=
[[[[
[
0 0
11965 10297
0 0
10297 191501
]]]]
]
119875 =
[[[[
[
162602 minus06640 10734 minus00095
minus06640 03007 minus03455 02871
10734 minus03455 04730 minus03458
minus00095 02871 minus03458 03658
]]]]
]
1198701= [
minus04509 01745 minus07182 00527
minus02807 00034 minus00471 02639]
1198702= [
minus04619 01238 minus07812 00351
minus02827 00068 minus00851 01401]
1198703= [
minus05902 02132 minus08910 00531
minus04107 00117 minus00730 02989]
1198704= [
minus04891 02109 minus06734 004145
minus03180 00085 minus00631 03893]
(36)
which guarantee the stability condition (17) MATLAB Tool-box is used to obtain parameters as 119896
1= 00251 and
8 The Scientific World Journal
Table 2 Peak-peak error and phase lag for Figure 5
The proposed approach PDCPeak-peak error Phase lag Peak-peak error Phase lag
1205791
1205792
1205791
1205792
1205791
1205792
1205791
1205792
07 035 49∘ 45∘ 13 065 162∘ 108∘
20
10
0
minus10
minus20
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
ReferenceActual
15
0
minus15
1205792
(a) The proposed approach
20
10
0
minus10
minus20
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
ReferenceActual
15
0
minus15
1205792
(b) PDC
Figure 4 Sinusoidal wave response for knee and ankle joints
1198962= 001539 For comparison with the proposed controller
the PDC feedback gains are designed to be
1= [
minus03293 01023 minus02538 002317
minus01783 00021 minus002672 04732]
1198702= [
minus04278 01845 minus02451 00731
minus02185 00026 minus00752 00923]
1198703= [
minus06013 03461 minus05482 00231
minus04421 00093 minus00651 03529]
1198704= [
minus03756 03150 minus05391 00421
minus04250 00023 minus00531 03597]
(37)
51 Sinusoidal Wave Response Continuous reciprocation isrequired in order to foster the recovery of extremity functionThe sinusoidal wave responses of the proposed approach andPDC for both knee and ankle joints are shown in Figure 4It is evident that angle trajectories of the proposed approachare close to the command Figure 5 shows that the proposedapproach exhibits less tracking errors than does PDC Thepeak-peak error and phase lag are listed in Table 2 Because
of the interaction of the two joints PDC has significant angleerrors for 120579
1 which will degrade the rehabilitation effect
However supervisory control can overcome the couplingeffect of the two joints to achieve excellent rehabilitationfunction for patients
52 Irregular Curve Response In practical applications itcould be expected that the reference command will changewith different input frequencies The desired trajectories forboth knee and ankle joints are
1205791= 20 lowast 033 (sin (2120587119891
1119905) + sin (2120587119891
2119905) + sin (2120587119891
3119905))
1205792= 15 lowast 033 (sin (2120587119891
1119905) + sin (2120587119891
2119905) + sin (2120587119891
3119905))
(38)
with 1198911= 005Hz 119891
2= 01Hz and 119891
3= 0066Hz
Figure 6 shows the tracking responses of irregular curvesobtained using both the proposed approach and PDC Track-ing errors for the knee and ankle joints are shown in Figure 7Clearly the angle error of the proposed approach is averagemaintained within 2∘ However the proposed approach iscapable of adapting to different frequencies
The Scientific World Journal 9
14
12
10
8
6
4
2
0
minus2
minus41009080706050403020100
Time (s)
Ang
le (d
eg)
1205791 error1205792 error
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 5 Angle tracking errors for both knee and ankle
20
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
120579215
10
5
0
minus5
minus10
minus15
Actual
(a) The proposed approach
20
10
0
minus10
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
1205792
Actual
(b) PDC
Figure 6 Irregular curve response for both knee and ankle joints
53 Elliptic Response The desired end-effect or trajectory isgiven by
119909119889(119905) = 0614 minus 0015 sdot cos (02120587 sdot 119905 minus 120587)
119910119889(119905) = minus01 sdot sin (02120587 sdot 119905 minus 2120587)
(39)
where 0 le 119905 le 20 seconds
The end-effect tracking responses in the 119909 119910 coordinatefor both the proposed approach and PDC are shown inFigure 8 and the end-effect position tracking errors aredisplayed in Figure 9 It is evident that tracking behavior ofthe proposed approach is better than that of the PDC Ascan be seen the tracking errors of the proposed approach are
10 The Scientific World Journal
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 7 Angle tracking errors for both the proposed approach and PDC
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(a) The proposed approach
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(b) PDC
Figure 8 Elliptic response of the proposed approach and PDC
within 003m On the other hand angle tracking errors ofknee and ankle joints are shown in Figure 10
Moreover it is difficult to enhance end-effector trackingperformance using the PDC algorithm because the PDCcannot overcome the nonlinearity of PAMs and the structuralinteraction However the proposed approach overcomessuccessfully the coupling effect and parameter uncertaintiesof the system As seen in the experimental results theproposed approach can attain excellent end-effector trackingperformance in rehabilitation function
6 Conclusions
In this study a novel composite fuzzy control is proposed andapplied in the two-joint leg rehabilitation device driven byPAMsThe proposed controller is not only capable of decom-posing nonlinear systems into a set of linear subsystems but isalso capable of simplifying a complex nonlinear system usinglinear control techniques with the control gains determinedusingMATLABrsquos LMI Toolbox based on the Lyapunov stabil-ity theoremMoreover the supervisory control can overcome
The Scientific World Journal 11
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(a) The proposed approach
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(b) PDC
Figure 9 The end-effect position tracking errors
14
12
10
8
6
4
2
0
minus2
minus40 2 4 6 8 10 12 14 16 18 20
Ang
le (d
eg)
Time (s)
1205791 error1205792 error
(a) The proposed approach
0 2 4 6 8 10 12 14 16 18 20
Time (s)
15
10
5
0
minus5
minus10
1205791 error1205792 error
Ang
le (d
eg)
(b) PDC
Figure 10 Angle tracking errors of knee and ankle joints
the coupling effect for a leg rehabilitation machine Experi-mental results show that the system response of the proposedapproach was in good agreement with that of the refer-ence input and guarantee robustness to system parameteruncertainties
Conflict of Interests
The author declares that there is no conflict of interestsregarding the publication of this paper
References
[1] D G Caldwell G A Medrano-Cerda and M GoodwinldquoControl of pneumatic muscle actuatorsrdquo IEEE Control SystemsMagazine vol 15 no 1 pp 40ndash48 1995
[2] C P Chou and B Hannaford ldquoStatic and dynamic characteris-tics ofMcKibben pneumatic artificialmusclesrdquo inProceedings ofthe IEEE International Conference on Robotics and Automationpp 281ndash286 May 1994
[3] T Noritsugu and T Tanaka ldquoApplication of rubber artificialmuscle manipulator as a rehabilitation robotrdquo IEEEASMETransactions on Mechatronics vol 2 no 4 pp 259ndash267 1997
12 The Scientific World Journal
[4] J H Lilly and L Yang ldquoSliding mode tracking for pneumaticmuscle actuators in opposing pair configurationrdquo IEEE Trans-actions on Control Systems Technology vol 4 pp 550ndash558 2005
[5] K K Ahn and H P H Anh ldquoDesign and implementation of anadaptive recurrent neural networks (ARNN) controller of thepneumatic artificial muscle (PAM) manipulatorrdquoMechatronicsvol 19 no 6 pp 816ndash828 2009
[6] X Shen ldquoNonlinearmodel-based control of pneumatic artificialmuscle servo systemsrdquo Control Engineering Practice vol 18 no3 pp 311ndash317 2010
[7] L A Zadeh ldquoFuzzy setsrdquo Information and Control vol 8 no 3pp 338ndash353 1965
[8] T Takagi and M Sugeno ldquoFuzzy identification of systems andits applications to modeling and controlrdquo IEEE Transactions onSystems Man and Cybernetics vol 15 no 1 pp 116ndash132 1985
[9] K K Ahn and H P H Anh ldquoInverse double NARX fuzzymodeling for system identificationrdquo IEEEASME Transactionson Mechatronics vol 15 no 1 pp 136ndash148 2010
[10] L Seddiki K Guelton and J Zaytoon ldquoConcept and Takagi-Sugeno descriptor tracking controller design of a closedmuscu-lar chain lower-limb rehabilitation devicerdquo IET Control Theoryand Applications vol 4 no 8 pp 1407ndash1420 2010
[11] K Tanaka T Ikeda and H O Wang ldquoFuzzy regulators andfuzzy observers relaxed stability conditions and LMI-baseddesignsrdquo IEEE Transactions on Fuzzy Systems vol 6 no 2 pp250ndash265 1998
[12] M K Chang J J Liou andM L Chen ldquoT-S fuzzymodel-basedtracking control of a one-dimensional manipulator actuated bypneumatic artificial musclesrdquo Control Engineering Practice vol19 no 12 pp 1442ndash1449 2011
[13] D B Reynolds D W Repperger C A Phillips and G BandryldquoModeling the dynamic characteristics of pneumatic musclerdquoAnnals of Biomedical Engineering vol 31 no 3 pp 310ndash3172003
[14] C S Tseng B S Chen and H J Uang ldquoFuzzy tracking controldesign for nonlinear dynamic systems via T-S fuzzy modelrdquoIEEE Transactions on Fuzzy Systems vol 9 no 3 pp 381ndash3922001
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
The Scientific World Journal 3
controller 120591(119905) is designed on the basis of parallel distributedcompensation (PDC) and is represented as follows
Rule 119894
IF 1199111(119905) is 119865
1119894and sdot sdot sdot and 119911
119892(119905) is 119865
119892119894
THEN 120591 (119905) = minus119870119894119909 (119905)
(8)
where119870119894represents feedback gainThe inferred output of the
PDC controller is expressed in the following form
120591 (119905) = minus
119903
sum
119894=1
ℎ119894119870119894119909 (119905) (9)
Substituting (9) into (7) yields
119909 (119905) =
119903
sum
119894=1
119903
sum
119895=1
ℎ119894ℎ119895(119860119894minus 119861119894119870119895) 119909 (119905) (10)
The stability analysis of this tracking system (10) is carried outusing the Lyapunov direct method and the Lyapunov func-tion is defined as
119881 (119909 (119905)) = 119909119879
(119905) 119875119909 (119905) gt 0 (11)
where 119875 is a positive symmetric matrix Taking the derivativeof 119881 with respect to time yields
(119909 (119905)) = 119909119879
(119905) 119875119909 + 119909119879
(119905) 119875 119909 (119905)
=
119903
sum
119894=1
119903
sum
119895=1
ℎ119894ℎ119895(119909119879
(119860119879
119894minus 119870119879
119895119861119879
119894) 119875119909)
+
119903
sum
119894=1
119903
sum
119895=1
ℎ119894ℎ119895119909119879
119875 (119860119894minus 119861119894119870119895) 119909
=
119903
sum
119894=1
119903
sum
119895=1
ℎ119894ℎ119895119909119879
[(119860119894minus 119861119894119870119894)119879
119875 + 119875 (119860119894minus 119861119894119870119894)] 119909
(12)
The controller is stable if lt 0 Hence the LMI form isexpressed as follows
(119860119894minus 119861119894119870119894)119879
119875 + 119875 (119860119894minus 119861119894119870119894) lt 0 for 119894 = 1 2 119903
119866119894119895119875 + 119875119866
119894119895lt 0 1 le 119894 lt 119895 le 119903
(13)
where 119866119894119895= (119860119894minus 119861119894119870119895+ 119860119895minus 119861119895119870119894)2 and 119866
119894119894= 119860119894minus 119861119894119870119894
The controller gain 119870119894is obtained using the LMI toolbox
ofMATLAB If there exists a commonpositive definitematrix119875 that satisfies inequalities (13) it can be guaranteed that thetracking error will approach zero
22 Composite Fuzzy Tracking Controller Because the legrehabilitation machine has a coupling effect due to mecha-nism interaction many fuzzymodel controllers in the relatedliterature exhibit restrictive tracking control in application
The proposed approach introduces supervisory control inorder to overcome the coupling effect The 119894th rule of theproposed controller is defined as follows
Rule 119894
IF 1199111(119905) is 119865
1119894and sdot sdot sdot and 119911
119892(119905) is 119865
119892119894
THEN 119906 (119905) = 119870119894119909 (119905) + 119906
119904119894 = 1 2 119903
(14)
where 119906119904(119905) isin 119877
119898 The proposed controller consists of a localstate feedback 119870
119894119909 and a supervisory control 119906
119904 Therefore
the output of the proposed controller is
119906 (119905) =
119903
sum
119894
ℎ119894119870119894119909 (119905) + 119906
119904 (15)
The closed-loop system is given by
(119905) =
119903
sum
119894=1
119903
sum
119895=1
ℎ119894ℎ119895[119860119894minus 119861119894119870119895] 119909 (119905) +
119903
sum
119894=1
ℎ119894119861119894119906119904(119905)
=
119903
sum
119894=1
ℎ2
119894119866119894119894119909 (119905) + 2
119903
sum
119894lt119895
ℎ119894ℎ119895119866119894119895119909 (119905) + 119861119906
119904(119905)
(16)
Suppose that there exist a symmetric and positive definitematrix 119875 and some matrices119870
119894so that the following reduced
stability condition holds
(119860119894minus 119861119894119870119894)119879
119875 + 119875 (119860119894minus 119861119894119870119894) le minus119876
119894 119894 = 1 119903 (17)
where 119876119894is a positive definite matrix Based on this assump-
tion each subsystem is locally controllable and a stablefeedback gain is obtainable Intuitively a common matrix 119875
that satisfies (17) can be obtained more easily than can onethat fulfills the basic stabilization conditions When the LMImethod is applied conditions (17) can be efficiently verified Ifa feasible solution is obtained the design proceeds to exploitthe supervisory control in order to deal with the couplingeffects
Choose the Lyapunov function candidate 1198811(119909) = 119909
119879
119875119909The time derivative of 119881
1(119909) is as follows
1(119909) =
119903
sum
119894=1
ℎ2
119894119909119879
(119866119879
119894119894119875 + 119875119866
119894119894) 119909
+ 2
119903
sum
119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909 + 2119909
119879
119875119861119906119904
le minus
119903
sum
119894=1
ℎ2
119894119909119879
119876119894119909
+ 2
119903
sum
119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909 + 2119909
119879
119875119861119906119904
(18)
Given the matrix property clearly
120582min (119866119879
119894119895119875 + 119875119866
119894119895) 1199092
le 119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909
le 120582min (119866119879
119894119895119875 + 119875119866
119894119895) 1199092
(19)
4 The Scientific World Journal
where 120582min(max) denotes the smallest (largest) eigenvalue ofthe matrix Define
120572 = max119894119895
120582max (119866119879
119894119895119875 + 119875119866
119894119895) for 1 le 119894 lt 119895 le 119903 (20)
A relaxed condition concerning the coupling effect isexpressed as
119903
sum
119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909 le 119896
11199092
1198961=
119903 (119903 minus 1)
2120572
(21)
Finding the maximum value of sum119903119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895)119909
is equivalent to determining the maximum value ofsum119903
119894lt119895ℎ119894ℎ119895120582max(119866
119879
119894119895119875 + 119875119866
119894119895) This can be presented as
a nonlinear programming The optimal algorithms areemployed to seek the best solution Moreover the MATLABOptimization Toolbox consists of functions that minimize ormaximize general nonlinear functions By using the toolboxthe nonlinear programming is expressed in the followingform
max119894119895
119903
sum
119894lt119895
ℎ119894ℎ119895120582max (119866
119879
119894119895119875 + 119875119866
119894119895) 1 le 119894 lt 119895 le 119903
Subject to119903
sum
119894=1
120583119894= 1 120583
119894ge 0
119903
sum
119895=1
120583119895= 1 120583
119895ge 0
(22)
The largest eigenvalue of (119866119879119894119895119875 + 119875119866
119894119895) can be obtained in
advance so the maximum value is determined to be
1198962= max119894119895
sum
119894lt119895
ℎ119894ℎ119895120582max (119866
119879
119894119895119875 + 119875119866
119894119895) (23)
The following supervisory control is chosen
119906119904=
minus119861119879
119875119909
1003817100381710038171003817119909119879119875119861
1003817100381710038171003817
21198961199092
if 10038171003817100381710038171003817119909119879
11987511986110038171003817100381710038171003817
= 0
0 if 10038171003817100381710038171003817119909119879
11987511986110038171003817100381710038171003817= 0
(24)
where 119896 gt 119896119895 119895 = 1 or 2 If 119909119879119875119861 = 0 then substituting (24)
into (18) gives
1(119909) le minus
119903
sum
119894=1
ℎ2
119894119909119879
119876119894119909 + 2119896
1198951199092
minus 21198961199092
le minus
119903
sum
119894=1
ℎ2
119894119909119879
119876119894119909 = minus119881
2(119909)
(25)
where1198812(119909) is a positive definite functionWhen 119909
119879
119875119861 = 0
can give the following form
119909119879
[(119860119894minus 119861119894119870119894)119879
119875 + 119875 (119860119894minus 119861119894119870119894)] 119909
= 119909119879
(119860119879
119894119875 + 119875119860
119894) 119909 le minus119909
119879
119876119894119909 119894 = 1 119903
119903
sum
119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909
=
119903
sum
119894lt119895
ℎ119894ℎ119895[119909119879
(119860119894119875 minus 119875119860
119894) 119909 + 119909
119879
(119860119879
119895119875 minus 119875119860
119895) 119909]
le minussum
119894lt119895
ℎ119894ℎ119895119909119879
(119876119894+ 119876119891) 119909 1 le 119894 lt 119895 le 119903
(26)
the time derivative of 1198811(119909) becomes
1(119909) le minus
119903
sum
119894=1
ℎ2
119894119909119879
119876119894119909 minussum
119894lt119895
ℎ119894ℎ119895119909119879
(119876119894+ 119876119891) 119909 = minus119881
3(119909)
(27)
where 1198813(119909) is a positive definite function Thus the closed-
loop fuzzy system is asymptotically stable
3 System Descriptions
Figure 1 shows the experimental setup including four PAMstwo rotary potentiometers four pressure proportional valvesand four pressure transducers The hardware includes anIBM-compatible personal computer to calculate the con-trol signal which controls the pressure proportional valvethrough a DA card The angles of the joints are detectedusing rotary potentiometers the air pressure of each PAM ismeasured using pressure transducers and the measurementsare then fed back to the computer through anAD cardThesespecifications are listed in Table 1
Figure 2 presents the operation principle of the leg reha-bilitation machine depicting a two-joint leg The behavior ofthe leg manipulated by the rehabilitation machine is similarto that of a human leg Output angles 120579
1and 1205792simulate the
knee and ankle joints and the ranges of the rotary angles 1205791
and 1205792are fromminus45
∘ to 45∘ and fromminus50∘ to 50∘ respectively
The link mass 1198981= 27 kg119898
2= 081 kg and the link length
1198971= 05m 119897
2= 026m The rotating torque 120591 is generated
by the difference in pressure Δ119901 between the two opposingPAMs That is when 119901
119886gt 119901119887 as in Figure 2 the torque
exerted on the joint is counterclockwise and the rotation ofthe joint is also counterclockwise
So a pair of such PAMs is tied together around a pulleywith a radius 119903
119894 as in Figure 2 Then the torque values
imparted to the pulley by the PAM pair are [12]
1205911= (1206011119886
minus 1206011119887) 1199031
1205912= (1206012119886
minus 1206012119887) 1199032
(28)
The Scientific World Journal 5
1
2
3
4
5
6
78910
11
12
1314
15
16
ADDA
Figure 1 The experimental setup
1b
1205911
2b
1205912 m2
r1
1205792
2a
1a
m1
r1
1205791
l1
l2
p1b = p10 minus Δp1
p1a = p10 + Δp1
p2a=p 20
+ Δp2
p 2b=p 20
minus Δp2
Figure 2 Operation principle of the leg rehabilitation machine driven by PAMs
6 The Scientific World Journal
Table 1 Component specifications
Number Component Specifications1 2 3 4 PAM Festo MAS-20-150N7 8 9 10 Pressure proportional valve Mac PPC5C5 6 Rotary potentiometer Keen Engineering KRT205011 12 13 14 Pressure transducer Jihsense SN 911166 0ndash10 kgfcm2
16 IBM-compatible PC Pentium 18GHzDAC ADC Automation AIO3321
where
1206011119886
= 1198651119886(1199011119886) minus 1198701119886(1199011119886) 11990311205791minus 1198611119886(1199011119886) 1199031
1205791
(29a)
1206011119887
= 1198651119887(1199011119887) minus 1198701119887(1199011119887) 11990311205791minus 1198611119887(1199011119887) 1199031
1205791
(29b)
1206012119886
= 1198652119886(1199012119886) minus 1198702119886(1199012119886) 11990321205792minus 1198612119886(1199012119886) 1199032
1205792
(29c)
1206012119887
= 1198652119887(1199012119887) minus 1198702119887(1199012119887) 11990321205792minus 1198612119887(1199012119887) 1199032
1205792 (29d)
where the spring coefficient119870(119901) and the damping coefficient119861(119901) are given by Reynolds et al [13]
The desired input pressures P119886= [1199011119886
1199012119886]119879 and P
119887=
[1199011119887
1199012119887]119879 for each PAM are generated by the following
equation
P119886(119905) = P
0+ ΔP (119905) P
119887(119905) = P
0minus ΔP (119905) (30)
where P0
= [11990110
11990120]119879 is a nominal constant input PAM
pressure and ΔP(119905) = [Δ1199011
Δ1199012]119879 is the control pressure
input with an arbitrary function of time Because the pressureinputΔP(119905) = [Δ119901
1Δ1199012]119879 and output 120579 = [120579
11205792]119879 the sys-
tem can bewritten as a two-input two-output (TITO) controlsystem The control signal u = [119906
11199062]119879 is proportional to
ΔP based on the pressure proportional valversquos characteristicsThat is ΔP can be used instead of u as a control input
4 Dynamic Model of a Two-Joint LegRehabilitation Machine Driven by PAMs
Figure 2 shows a two-joint leg rehabilitation machine drivenby PAMs and the dynamic equation is given as follows [14]
119872(120579) 120579 + 119862 (120579 120579) 120579 + 119866 (120579) = 120591 (31)
where
119872(120579) = [(1198981+ 1198982) 1198972
119898211989711198972(11990411199042+ 11988811198882)
119898211989711198972(11990411199042+ 11988811198882) 119898
21198972
2
]
119862 (120579 120579) = [0 minus119898
211989711198972(11988811199042minus 11990411198882) 1205792
minus119898211989711198972(11988811199042minus 11990411198882) 1205791
0]
119866 (120579) = [(1198981+ 1198982) 11989711198921199041
minus119898211989721198921199042
]
(32)
and119872(120579) is the moment of inertia 119862(120579 120579) includes Coriolisand centripetal force and 119866(120579) is the gravitational force
Notation 1199041= sin(120579
1) 1199042= sin(120579
2) 1198881= cos(120579
1) and 119888
2=
cos(1205792) Let 119909
1= 1205791 1199092= 1205791 1199093= 1205792 and 119909
4= 1205792 then (31)
can be written as the following state-space form [14]
1= 1199092
2= 1198911(119909) + 119892
11(119909) 1205911+ 119892121205912
3= 1199094
4= 1198912(119909) + 119892
21(119909) + 119892
221205912
(33)
where1198911(119909)
=
(11990411198882minus 11988811199042) [119898211989711198972(11990411199042+ 11988811198882) 1199092
2minus 11989821198972
21199092
4]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
+[(1198981minus 1198982) 11989721198921199041minus 119898211989721198921199042(11990411199042+ 11988811198882)]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
1198912(119909)
=
(11990411198882minus 11988811199042) [minus (119898
1+ 1198982) 1198972
11199092
2+ 119898211989711198972(11990411199042+ 11988811198882) 1199092
4]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
+[minus (119898
1+ 1198982) 11989711198921199041(11990411199042+ 11988811198882) + (119898
1+ 1198982) 11989711198921199042]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989211(119909) =
11989821198972
2
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989212(119909) =
minus119898211989711198972(11990411199042+ 11988811198882)
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989221(119909) =
minus119898211989711198972(11990411199042+ 11988811198882)
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989222(119909) =
(1198981+ 1198982) 1198972
1
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
(34)
5 Experimental Studies
Figure 3 shows the leg rehabilitation machine using an actualhuman loading with a 65 kg weight The automatic device
The Scientific World Journal 7
Figure 3 The two-joint leg rehabilitation device with actual human loading
can help patients to recover lower limb motion function bymeans of continuous passive motion such as a sinusoidalwave command an irregular curve command and an end-effect tracking command The experiments include both theproposed approach and PDC for comparison in order toevaluate efficacy and control performance The controllerswere implemented on an Intel Pentium 18GHz PC with asampling time of 5ms and the entire control software wascoded in C++
This study attempts to use as few rules as possible inorder tominimize design effort and complexityTheT-S fuzzymodel of the system is thus given the following four-rulefuzzy model
1198771
IF 1199091is about (1205874) and 119909
3is about (1205874)
THEN = 11986011199091+ 1198611119906
1198772
IF 1199091is about (1205874) and 119909
3is about (minus1205874)
THEN = 11986021199091+ 1198612119906
1198773
IF 1199091is about (minus1205874) and 119909
3is about (1205874)
THEN = 11986031199091+ 1198613119906
1198774
IF 1199091is about (minus1205874) and 119909
3is about (minus1205874)
THEN = 11986041199091+ 1198614119906
(35)
where
1198601=
[[[
[
0 1 0 0
44082 00059 06742 minus00002
0 0 0 1
minus14572 00002 minus39722 00002
]]]
]
1198602=
[[[
[
0 1 0 0
46734 00049 0532 00001
0 0 0 1
minus12145 minus00001 minus3563 00001
]]]
]
1198603=
[[[
[
0 1 0 0
54782 00021 06876 minus00002
0 0 0 1
14525 00002 46723 00002
]]]
]
1198604=
[[[[
[
0 1 0 0
49821 minus00044 06543 00002
0 0 0 1
minus11034 minus00002 35631 00003
]]]]
]
1198611= 1198614=
[[[[
[
0 0
14811 minus2849
0 0
minus2849 237417
]]]]
]
1198612= 1198613=
[[[[
[
0 0
11965 10297
0 0
10297 191501
]]]]
]
119875 =
[[[[
[
162602 minus06640 10734 minus00095
minus06640 03007 minus03455 02871
10734 minus03455 04730 minus03458
minus00095 02871 minus03458 03658
]]]]
]
1198701= [
minus04509 01745 minus07182 00527
minus02807 00034 minus00471 02639]
1198702= [
minus04619 01238 minus07812 00351
minus02827 00068 minus00851 01401]
1198703= [
minus05902 02132 minus08910 00531
minus04107 00117 minus00730 02989]
1198704= [
minus04891 02109 minus06734 004145
minus03180 00085 minus00631 03893]
(36)
which guarantee the stability condition (17) MATLAB Tool-box is used to obtain parameters as 119896
1= 00251 and
8 The Scientific World Journal
Table 2 Peak-peak error and phase lag for Figure 5
The proposed approach PDCPeak-peak error Phase lag Peak-peak error Phase lag
1205791
1205792
1205791
1205792
1205791
1205792
1205791
1205792
07 035 49∘ 45∘ 13 065 162∘ 108∘
20
10
0
minus10
minus20
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
ReferenceActual
15
0
minus15
1205792
(a) The proposed approach
20
10
0
minus10
minus20
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
ReferenceActual
15
0
minus15
1205792
(b) PDC
Figure 4 Sinusoidal wave response for knee and ankle joints
1198962= 001539 For comparison with the proposed controller
the PDC feedback gains are designed to be
1= [
minus03293 01023 minus02538 002317
minus01783 00021 minus002672 04732]
1198702= [
minus04278 01845 minus02451 00731
minus02185 00026 minus00752 00923]
1198703= [
minus06013 03461 minus05482 00231
minus04421 00093 minus00651 03529]
1198704= [
minus03756 03150 minus05391 00421
minus04250 00023 minus00531 03597]
(37)
51 Sinusoidal Wave Response Continuous reciprocation isrequired in order to foster the recovery of extremity functionThe sinusoidal wave responses of the proposed approach andPDC for both knee and ankle joints are shown in Figure 4It is evident that angle trajectories of the proposed approachare close to the command Figure 5 shows that the proposedapproach exhibits less tracking errors than does PDC Thepeak-peak error and phase lag are listed in Table 2 Because
of the interaction of the two joints PDC has significant angleerrors for 120579
1 which will degrade the rehabilitation effect
However supervisory control can overcome the couplingeffect of the two joints to achieve excellent rehabilitationfunction for patients
52 Irregular Curve Response In practical applications itcould be expected that the reference command will changewith different input frequencies The desired trajectories forboth knee and ankle joints are
1205791= 20 lowast 033 (sin (2120587119891
1119905) + sin (2120587119891
2119905) + sin (2120587119891
3119905))
1205792= 15 lowast 033 (sin (2120587119891
1119905) + sin (2120587119891
2119905) + sin (2120587119891
3119905))
(38)
with 1198911= 005Hz 119891
2= 01Hz and 119891
3= 0066Hz
Figure 6 shows the tracking responses of irregular curvesobtained using both the proposed approach and PDC Track-ing errors for the knee and ankle joints are shown in Figure 7Clearly the angle error of the proposed approach is averagemaintained within 2∘ However the proposed approach iscapable of adapting to different frequencies
The Scientific World Journal 9
14
12
10
8
6
4
2
0
minus2
minus41009080706050403020100
Time (s)
Ang
le (d
eg)
1205791 error1205792 error
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 5 Angle tracking errors for both knee and ankle
20
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
120579215
10
5
0
minus5
minus10
minus15
Actual
(a) The proposed approach
20
10
0
minus10
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
1205792
Actual
(b) PDC
Figure 6 Irregular curve response for both knee and ankle joints
53 Elliptic Response The desired end-effect or trajectory isgiven by
119909119889(119905) = 0614 minus 0015 sdot cos (02120587 sdot 119905 minus 120587)
119910119889(119905) = minus01 sdot sin (02120587 sdot 119905 minus 2120587)
(39)
where 0 le 119905 le 20 seconds
The end-effect tracking responses in the 119909 119910 coordinatefor both the proposed approach and PDC are shown inFigure 8 and the end-effect position tracking errors aredisplayed in Figure 9 It is evident that tracking behavior ofthe proposed approach is better than that of the PDC Ascan be seen the tracking errors of the proposed approach are
10 The Scientific World Journal
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 7 Angle tracking errors for both the proposed approach and PDC
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(a) The proposed approach
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(b) PDC
Figure 8 Elliptic response of the proposed approach and PDC
within 003m On the other hand angle tracking errors ofknee and ankle joints are shown in Figure 10
Moreover it is difficult to enhance end-effector trackingperformance using the PDC algorithm because the PDCcannot overcome the nonlinearity of PAMs and the structuralinteraction However the proposed approach overcomessuccessfully the coupling effect and parameter uncertaintiesof the system As seen in the experimental results theproposed approach can attain excellent end-effector trackingperformance in rehabilitation function
6 Conclusions
In this study a novel composite fuzzy control is proposed andapplied in the two-joint leg rehabilitation device driven byPAMsThe proposed controller is not only capable of decom-posing nonlinear systems into a set of linear subsystems but isalso capable of simplifying a complex nonlinear system usinglinear control techniques with the control gains determinedusingMATLABrsquos LMI Toolbox based on the Lyapunov stabil-ity theoremMoreover the supervisory control can overcome
The Scientific World Journal 11
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(a) The proposed approach
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(b) PDC
Figure 9 The end-effect position tracking errors
14
12
10
8
6
4
2
0
minus2
minus40 2 4 6 8 10 12 14 16 18 20
Ang
le (d
eg)
Time (s)
1205791 error1205792 error
(a) The proposed approach
0 2 4 6 8 10 12 14 16 18 20
Time (s)
15
10
5
0
minus5
minus10
1205791 error1205792 error
Ang
le (d
eg)
(b) PDC
Figure 10 Angle tracking errors of knee and ankle joints
the coupling effect for a leg rehabilitation machine Experi-mental results show that the system response of the proposedapproach was in good agreement with that of the refer-ence input and guarantee robustness to system parameteruncertainties
Conflict of Interests
The author declares that there is no conflict of interestsregarding the publication of this paper
References
[1] D G Caldwell G A Medrano-Cerda and M GoodwinldquoControl of pneumatic muscle actuatorsrdquo IEEE Control SystemsMagazine vol 15 no 1 pp 40ndash48 1995
[2] C P Chou and B Hannaford ldquoStatic and dynamic characteris-tics ofMcKibben pneumatic artificialmusclesrdquo inProceedings ofthe IEEE International Conference on Robotics and Automationpp 281ndash286 May 1994
[3] T Noritsugu and T Tanaka ldquoApplication of rubber artificialmuscle manipulator as a rehabilitation robotrdquo IEEEASMETransactions on Mechatronics vol 2 no 4 pp 259ndash267 1997
12 The Scientific World Journal
[4] J H Lilly and L Yang ldquoSliding mode tracking for pneumaticmuscle actuators in opposing pair configurationrdquo IEEE Trans-actions on Control Systems Technology vol 4 pp 550ndash558 2005
[5] K K Ahn and H P H Anh ldquoDesign and implementation of anadaptive recurrent neural networks (ARNN) controller of thepneumatic artificial muscle (PAM) manipulatorrdquoMechatronicsvol 19 no 6 pp 816ndash828 2009
[6] X Shen ldquoNonlinearmodel-based control of pneumatic artificialmuscle servo systemsrdquo Control Engineering Practice vol 18 no3 pp 311ndash317 2010
[7] L A Zadeh ldquoFuzzy setsrdquo Information and Control vol 8 no 3pp 338ndash353 1965
[8] T Takagi and M Sugeno ldquoFuzzy identification of systems andits applications to modeling and controlrdquo IEEE Transactions onSystems Man and Cybernetics vol 15 no 1 pp 116ndash132 1985
[9] K K Ahn and H P H Anh ldquoInverse double NARX fuzzymodeling for system identificationrdquo IEEEASME Transactionson Mechatronics vol 15 no 1 pp 136ndash148 2010
[10] L Seddiki K Guelton and J Zaytoon ldquoConcept and Takagi-Sugeno descriptor tracking controller design of a closedmuscu-lar chain lower-limb rehabilitation devicerdquo IET Control Theoryand Applications vol 4 no 8 pp 1407ndash1420 2010
[11] K Tanaka T Ikeda and H O Wang ldquoFuzzy regulators andfuzzy observers relaxed stability conditions and LMI-baseddesignsrdquo IEEE Transactions on Fuzzy Systems vol 6 no 2 pp250ndash265 1998
[12] M K Chang J J Liou andM L Chen ldquoT-S fuzzymodel-basedtracking control of a one-dimensional manipulator actuated bypneumatic artificial musclesrdquo Control Engineering Practice vol19 no 12 pp 1442ndash1449 2011
[13] D B Reynolds D W Repperger C A Phillips and G BandryldquoModeling the dynamic characteristics of pneumatic musclerdquoAnnals of Biomedical Engineering vol 31 no 3 pp 310ndash3172003
[14] C S Tseng B S Chen and H J Uang ldquoFuzzy tracking controldesign for nonlinear dynamic systems via T-S fuzzy modelrdquoIEEE Transactions on Fuzzy Systems vol 9 no 3 pp 381ndash3922001
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
4 The Scientific World Journal
where 120582min(max) denotes the smallest (largest) eigenvalue ofthe matrix Define
120572 = max119894119895
120582max (119866119879
119894119895119875 + 119875119866
119894119895) for 1 le 119894 lt 119895 le 119903 (20)
A relaxed condition concerning the coupling effect isexpressed as
119903
sum
119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909 le 119896
11199092
1198961=
119903 (119903 minus 1)
2120572
(21)
Finding the maximum value of sum119903119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895)119909
is equivalent to determining the maximum value ofsum119903
119894lt119895ℎ119894ℎ119895120582max(119866
119879
119894119895119875 + 119875119866
119894119895) This can be presented as
a nonlinear programming The optimal algorithms areemployed to seek the best solution Moreover the MATLABOptimization Toolbox consists of functions that minimize ormaximize general nonlinear functions By using the toolboxthe nonlinear programming is expressed in the followingform
max119894119895
119903
sum
119894lt119895
ℎ119894ℎ119895120582max (119866
119879
119894119895119875 + 119875119866
119894119895) 1 le 119894 lt 119895 le 119903
Subject to119903
sum
119894=1
120583119894= 1 120583
119894ge 0
119903
sum
119895=1
120583119895= 1 120583
119895ge 0
(22)
The largest eigenvalue of (119866119879119894119895119875 + 119875119866
119894119895) can be obtained in
advance so the maximum value is determined to be
1198962= max119894119895
sum
119894lt119895
ℎ119894ℎ119895120582max (119866
119879
119894119895119875 + 119875119866
119894119895) (23)
The following supervisory control is chosen
119906119904=
minus119861119879
119875119909
1003817100381710038171003817119909119879119875119861
1003817100381710038171003817
21198961199092
if 10038171003817100381710038171003817119909119879
11987511986110038171003817100381710038171003817
= 0
0 if 10038171003817100381710038171003817119909119879
11987511986110038171003817100381710038171003817= 0
(24)
where 119896 gt 119896119895 119895 = 1 or 2 If 119909119879119875119861 = 0 then substituting (24)
into (18) gives
1(119909) le minus
119903
sum
119894=1
ℎ2
119894119909119879
119876119894119909 + 2119896
1198951199092
minus 21198961199092
le minus
119903
sum
119894=1
ℎ2
119894119909119879
119876119894119909 = minus119881
2(119909)
(25)
where1198812(119909) is a positive definite functionWhen 119909
119879
119875119861 = 0
can give the following form
119909119879
[(119860119894minus 119861119894119870119894)119879
119875 + 119875 (119860119894minus 119861119894119870119894)] 119909
= 119909119879
(119860119879
119894119875 + 119875119860
119894) 119909 le minus119909
119879
119876119894119909 119894 = 1 119903
119903
sum
119894lt119895
ℎ119894ℎ119895119909119879
(119866119879
119894119895119875 + 119875119866
119894119895) 119909
=
119903
sum
119894lt119895
ℎ119894ℎ119895[119909119879
(119860119894119875 minus 119875119860
119894) 119909 + 119909
119879
(119860119879
119895119875 minus 119875119860
119895) 119909]
le minussum
119894lt119895
ℎ119894ℎ119895119909119879
(119876119894+ 119876119891) 119909 1 le 119894 lt 119895 le 119903
(26)
the time derivative of 1198811(119909) becomes
1(119909) le minus
119903
sum
119894=1
ℎ2
119894119909119879
119876119894119909 minussum
119894lt119895
ℎ119894ℎ119895119909119879
(119876119894+ 119876119891) 119909 = minus119881
3(119909)
(27)
where 1198813(119909) is a positive definite function Thus the closed-
loop fuzzy system is asymptotically stable
3 System Descriptions
Figure 1 shows the experimental setup including four PAMstwo rotary potentiometers four pressure proportional valvesand four pressure transducers The hardware includes anIBM-compatible personal computer to calculate the con-trol signal which controls the pressure proportional valvethrough a DA card The angles of the joints are detectedusing rotary potentiometers the air pressure of each PAM ismeasured using pressure transducers and the measurementsare then fed back to the computer through anAD cardThesespecifications are listed in Table 1
Figure 2 presents the operation principle of the leg reha-bilitation machine depicting a two-joint leg The behavior ofthe leg manipulated by the rehabilitation machine is similarto that of a human leg Output angles 120579
1and 1205792simulate the
knee and ankle joints and the ranges of the rotary angles 1205791
and 1205792are fromminus45
∘ to 45∘ and fromminus50∘ to 50∘ respectively
The link mass 1198981= 27 kg119898
2= 081 kg and the link length
1198971= 05m 119897
2= 026m The rotating torque 120591 is generated
by the difference in pressure Δ119901 between the two opposingPAMs That is when 119901
119886gt 119901119887 as in Figure 2 the torque
exerted on the joint is counterclockwise and the rotation ofthe joint is also counterclockwise
So a pair of such PAMs is tied together around a pulleywith a radius 119903
119894 as in Figure 2 Then the torque values
imparted to the pulley by the PAM pair are [12]
1205911= (1206011119886
minus 1206011119887) 1199031
1205912= (1206012119886
minus 1206012119887) 1199032
(28)
The Scientific World Journal 5
1
2
3
4
5
6
78910
11
12
1314
15
16
ADDA
Figure 1 The experimental setup
1b
1205911
2b
1205912 m2
r1
1205792
2a
1a
m1
r1
1205791
l1
l2
p1b = p10 minus Δp1
p1a = p10 + Δp1
p2a=p 20
+ Δp2
p 2b=p 20
minus Δp2
Figure 2 Operation principle of the leg rehabilitation machine driven by PAMs
6 The Scientific World Journal
Table 1 Component specifications
Number Component Specifications1 2 3 4 PAM Festo MAS-20-150N7 8 9 10 Pressure proportional valve Mac PPC5C5 6 Rotary potentiometer Keen Engineering KRT205011 12 13 14 Pressure transducer Jihsense SN 911166 0ndash10 kgfcm2
16 IBM-compatible PC Pentium 18GHzDAC ADC Automation AIO3321
where
1206011119886
= 1198651119886(1199011119886) minus 1198701119886(1199011119886) 11990311205791minus 1198611119886(1199011119886) 1199031
1205791
(29a)
1206011119887
= 1198651119887(1199011119887) minus 1198701119887(1199011119887) 11990311205791minus 1198611119887(1199011119887) 1199031
1205791
(29b)
1206012119886
= 1198652119886(1199012119886) minus 1198702119886(1199012119886) 11990321205792minus 1198612119886(1199012119886) 1199032
1205792
(29c)
1206012119887
= 1198652119887(1199012119887) minus 1198702119887(1199012119887) 11990321205792minus 1198612119887(1199012119887) 1199032
1205792 (29d)
where the spring coefficient119870(119901) and the damping coefficient119861(119901) are given by Reynolds et al [13]
The desired input pressures P119886= [1199011119886
1199012119886]119879 and P
119887=
[1199011119887
1199012119887]119879 for each PAM are generated by the following
equation
P119886(119905) = P
0+ ΔP (119905) P
119887(119905) = P
0minus ΔP (119905) (30)
where P0
= [11990110
11990120]119879 is a nominal constant input PAM
pressure and ΔP(119905) = [Δ1199011
Δ1199012]119879 is the control pressure
input with an arbitrary function of time Because the pressureinputΔP(119905) = [Δ119901
1Δ1199012]119879 and output 120579 = [120579
11205792]119879 the sys-
tem can bewritten as a two-input two-output (TITO) controlsystem The control signal u = [119906
11199062]119879 is proportional to
ΔP based on the pressure proportional valversquos characteristicsThat is ΔP can be used instead of u as a control input
4 Dynamic Model of a Two-Joint LegRehabilitation Machine Driven by PAMs
Figure 2 shows a two-joint leg rehabilitation machine drivenby PAMs and the dynamic equation is given as follows [14]
119872(120579) 120579 + 119862 (120579 120579) 120579 + 119866 (120579) = 120591 (31)
where
119872(120579) = [(1198981+ 1198982) 1198972
119898211989711198972(11990411199042+ 11988811198882)
119898211989711198972(11990411199042+ 11988811198882) 119898
21198972
2
]
119862 (120579 120579) = [0 minus119898
211989711198972(11988811199042minus 11990411198882) 1205792
minus119898211989711198972(11988811199042minus 11990411198882) 1205791
0]
119866 (120579) = [(1198981+ 1198982) 11989711198921199041
minus119898211989721198921199042
]
(32)
and119872(120579) is the moment of inertia 119862(120579 120579) includes Coriolisand centripetal force and 119866(120579) is the gravitational force
Notation 1199041= sin(120579
1) 1199042= sin(120579
2) 1198881= cos(120579
1) and 119888
2=
cos(1205792) Let 119909
1= 1205791 1199092= 1205791 1199093= 1205792 and 119909
4= 1205792 then (31)
can be written as the following state-space form [14]
1= 1199092
2= 1198911(119909) + 119892
11(119909) 1205911+ 119892121205912
3= 1199094
4= 1198912(119909) + 119892
21(119909) + 119892
221205912
(33)
where1198911(119909)
=
(11990411198882minus 11988811199042) [119898211989711198972(11990411199042+ 11988811198882) 1199092
2minus 11989821198972
21199092
4]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
+[(1198981minus 1198982) 11989721198921199041minus 119898211989721198921199042(11990411199042+ 11988811198882)]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
1198912(119909)
=
(11990411198882minus 11988811199042) [minus (119898
1+ 1198982) 1198972
11199092
2+ 119898211989711198972(11990411199042+ 11988811198882) 1199092
4]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
+[minus (119898
1+ 1198982) 11989711198921199041(11990411199042+ 11988811198882) + (119898
1+ 1198982) 11989711198921199042]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989211(119909) =
11989821198972
2
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989212(119909) =
minus119898211989711198972(11990411199042+ 11988811198882)
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989221(119909) =
minus119898211989711198972(11990411199042+ 11988811198882)
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989222(119909) =
(1198981+ 1198982) 1198972
1
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
(34)
5 Experimental Studies
Figure 3 shows the leg rehabilitation machine using an actualhuman loading with a 65 kg weight The automatic device
The Scientific World Journal 7
Figure 3 The two-joint leg rehabilitation device with actual human loading
can help patients to recover lower limb motion function bymeans of continuous passive motion such as a sinusoidalwave command an irregular curve command and an end-effect tracking command The experiments include both theproposed approach and PDC for comparison in order toevaluate efficacy and control performance The controllerswere implemented on an Intel Pentium 18GHz PC with asampling time of 5ms and the entire control software wascoded in C++
This study attempts to use as few rules as possible inorder tominimize design effort and complexityTheT-S fuzzymodel of the system is thus given the following four-rulefuzzy model
1198771
IF 1199091is about (1205874) and 119909
3is about (1205874)
THEN = 11986011199091+ 1198611119906
1198772
IF 1199091is about (1205874) and 119909
3is about (minus1205874)
THEN = 11986021199091+ 1198612119906
1198773
IF 1199091is about (minus1205874) and 119909
3is about (1205874)
THEN = 11986031199091+ 1198613119906
1198774
IF 1199091is about (minus1205874) and 119909
3is about (minus1205874)
THEN = 11986041199091+ 1198614119906
(35)
where
1198601=
[[[
[
0 1 0 0
44082 00059 06742 minus00002
0 0 0 1
minus14572 00002 minus39722 00002
]]]
]
1198602=
[[[
[
0 1 0 0
46734 00049 0532 00001
0 0 0 1
minus12145 minus00001 minus3563 00001
]]]
]
1198603=
[[[
[
0 1 0 0
54782 00021 06876 minus00002
0 0 0 1
14525 00002 46723 00002
]]]
]
1198604=
[[[[
[
0 1 0 0
49821 minus00044 06543 00002
0 0 0 1
minus11034 minus00002 35631 00003
]]]]
]
1198611= 1198614=
[[[[
[
0 0
14811 minus2849
0 0
minus2849 237417
]]]]
]
1198612= 1198613=
[[[[
[
0 0
11965 10297
0 0
10297 191501
]]]]
]
119875 =
[[[[
[
162602 minus06640 10734 minus00095
minus06640 03007 minus03455 02871
10734 minus03455 04730 minus03458
minus00095 02871 minus03458 03658
]]]]
]
1198701= [
minus04509 01745 minus07182 00527
minus02807 00034 minus00471 02639]
1198702= [
minus04619 01238 minus07812 00351
minus02827 00068 minus00851 01401]
1198703= [
minus05902 02132 minus08910 00531
minus04107 00117 minus00730 02989]
1198704= [
minus04891 02109 minus06734 004145
minus03180 00085 minus00631 03893]
(36)
which guarantee the stability condition (17) MATLAB Tool-box is used to obtain parameters as 119896
1= 00251 and
8 The Scientific World Journal
Table 2 Peak-peak error and phase lag for Figure 5
The proposed approach PDCPeak-peak error Phase lag Peak-peak error Phase lag
1205791
1205792
1205791
1205792
1205791
1205792
1205791
1205792
07 035 49∘ 45∘ 13 065 162∘ 108∘
20
10
0
minus10
minus20
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
ReferenceActual
15
0
minus15
1205792
(a) The proposed approach
20
10
0
minus10
minus20
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
ReferenceActual
15
0
minus15
1205792
(b) PDC
Figure 4 Sinusoidal wave response for knee and ankle joints
1198962= 001539 For comparison with the proposed controller
the PDC feedback gains are designed to be
1= [
minus03293 01023 minus02538 002317
minus01783 00021 minus002672 04732]
1198702= [
minus04278 01845 minus02451 00731
minus02185 00026 minus00752 00923]
1198703= [
minus06013 03461 minus05482 00231
minus04421 00093 minus00651 03529]
1198704= [
minus03756 03150 minus05391 00421
minus04250 00023 minus00531 03597]
(37)
51 Sinusoidal Wave Response Continuous reciprocation isrequired in order to foster the recovery of extremity functionThe sinusoidal wave responses of the proposed approach andPDC for both knee and ankle joints are shown in Figure 4It is evident that angle trajectories of the proposed approachare close to the command Figure 5 shows that the proposedapproach exhibits less tracking errors than does PDC Thepeak-peak error and phase lag are listed in Table 2 Because
of the interaction of the two joints PDC has significant angleerrors for 120579
1 which will degrade the rehabilitation effect
However supervisory control can overcome the couplingeffect of the two joints to achieve excellent rehabilitationfunction for patients
52 Irregular Curve Response In practical applications itcould be expected that the reference command will changewith different input frequencies The desired trajectories forboth knee and ankle joints are
1205791= 20 lowast 033 (sin (2120587119891
1119905) + sin (2120587119891
2119905) + sin (2120587119891
3119905))
1205792= 15 lowast 033 (sin (2120587119891
1119905) + sin (2120587119891
2119905) + sin (2120587119891
3119905))
(38)
with 1198911= 005Hz 119891
2= 01Hz and 119891
3= 0066Hz
Figure 6 shows the tracking responses of irregular curvesobtained using both the proposed approach and PDC Track-ing errors for the knee and ankle joints are shown in Figure 7Clearly the angle error of the proposed approach is averagemaintained within 2∘ However the proposed approach iscapable of adapting to different frequencies
The Scientific World Journal 9
14
12
10
8
6
4
2
0
minus2
minus41009080706050403020100
Time (s)
Ang
le (d
eg)
1205791 error1205792 error
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 5 Angle tracking errors for both knee and ankle
20
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
120579215
10
5
0
minus5
minus10
minus15
Actual
(a) The proposed approach
20
10
0
minus10
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
1205792
Actual
(b) PDC
Figure 6 Irregular curve response for both knee and ankle joints
53 Elliptic Response The desired end-effect or trajectory isgiven by
119909119889(119905) = 0614 minus 0015 sdot cos (02120587 sdot 119905 minus 120587)
119910119889(119905) = minus01 sdot sin (02120587 sdot 119905 minus 2120587)
(39)
where 0 le 119905 le 20 seconds
The end-effect tracking responses in the 119909 119910 coordinatefor both the proposed approach and PDC are shown inFigure 8 and the end-effect position tracking errors aredisplayed in Figure 9 It is evident that tracking behavior ofthe proposed approach is better than that of the PDC Ascan be seen the tracking errors of the proposed approach are
10 The Scientific World Journal
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 7 Angle tracking errors for both the proposed approach and PDC
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(a) The proposed approach
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(b) PDC
Figure 8 Elliptic response of the proposed approach and PDC
within 003m On the other hand angle tracking errors ofknee and ankle joints are shown in Figure 10
Moreover it is difficult to enhance end-effector trackingperformance using the PDC algorithm because the PDCcannot overcome the nonlinearity of PAMs and the structuralinteraction However the proposed approach overcomessuccessfully the coupling effect and parameter uncertaintiesof the system As seen in the experimental results theproposed approach can attain excellent end-effector trackingperformance in rehabilitation function
6 Conclusions
In this study a novel composite fuzzy control is proposed andapplied in the two-joint leg rehabilitation device driven byPAMsThe proposed controller is not only capable of decom-posing nonlinear systems into a set of linear subsystems but isalso capable of simplifying a complex nonlinear system usinglinear control techniques with the control gains determinedusingMATLABrsquos LMI Toolbox based on the Lyapunov stabil-ity theoremMoreover the supervisory control can overcome
The Scientific World Journal 11
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(a) The proposed approach
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(b) PDC
Figure 9 The end-effect position tracking errors
14
12
10
8
6
4
2
0
minus2
minus40 2 4 6 8 10 12 14 16 18 20
Ang
le (d
eg)
Time (s)
1205791 error1205792 error
(a) The proposed approach
0 2 4 6 8 10 12 14 16 18 20
Time (s)
15
10
5
0
minus5
minus10
1205791 error1205792 error
Ang
le (d
eg)
(b) PDC
Figure 10 Angle tracking errors of knee and ankle joints
the coupling effect for a leg rehabilitation machine Experi-mental results show that the system response of the proposedapproach was in good agreement with that of the refer-ence input and guarantee robustness to system parameteruncertainties
Conflict of Interests
The author declares that there is no conflict of interestsregarding the publication of this paper
References
[1] D G Caldwell G A Medrano-Cerda and M GoodwinldquoControl of pneumatic muscle actuatorsrdquo IEEE Control SystemsMagazine vol 15 no 1 pp 40ndash48 1995
[2] C P Chou and B Hannaford ldquoStatic and dynamic characteris-tics ofMcKibben pneumatic artificialmusclesrdquo inProceedings ofthe IEEE International Conference on Robotics and Automationpp 281ndash286 May 1994
[3] T Noritsugu and T Tanaka ldquoApplication of rubber artificialmuscle manipulator as a rehabilitation robotrdquo IEEEASMETransactions on Mechatronics vol 2 no 4 pp 259ndash267 1997
12 The Scientific World Journal
[4] J H Lilly and L Yang ldquoSliding mode tracking for pneumaticmuscle actuators in opposing pair configurationrdquo IEEE Trans-actions on Control Systems Technology vol 4 pp 550ndash558 2005
[5] K K Ahn and H P H Anh ldquoDesign and implementation of anadaptive recurrent neural networks (ARNN) controller of thepneumatic artificial muscle (PAM) manipulatorrdquoMechatronicsvol 19 no 6 pp 816ndash828 2009
[6] X Shen ldquoNonlinearmodel-based control of pneumatic artificialmuscle servo systemsrdquo Control Engineering Practice vol 18 no3 pp 311ndash317 2010
[7] L A Zadeh ldquoFuzzy setsrdquo Information and Control vol 8 no 3pp 338ndash353 1965
[8] T Takagi and M Sugeno ldquoFuzzy identification of systems andits applications to modeling and controlrdquo IEEE Transactions onSystems Man and Cybernetics vol 15 no 1 pp 116ndash132 1985
[9] K K Ahn and H P H Anh ldquoInverse double NARX fuzzymodeling for system identificationrdquo IEEEASME Transactionson Mechatronics vol 15 no 1 pp 136ndash148 2010
[10] L Seddiki K Guelton and J Zaytoon ldquoConcept and Takagi-Sugeno descriptor tracking controller design of a closedmuscu-lar chain lower-limb rehabilitation devicerdquo IET Control Theoryand Applications vol 4 no 8 pp 1407ndash1420 2010
[11] K Tanaka T Ikeda and H O Wang ldquoFuzzy regulators andfuzzy observers relaxed stability conditions and LMI-baseddesignsrdquo IEEE Transactions on Fuzzy Systems vol 6 no 2 pp250ndash265 1998
[12] M K Chang J J Liou andM L Chen ldquoT-S fuzzymodel-basedtracking control of a one-dimensional manipulator actuated bypneumatic artificial musclesrdquo Control Engineering Practice vol19 no 12 pp 1442ndash1449 2011
[13] D B Reynolds D W Repperger C A Phillips and G BandryldquoModeling the dynamic characteristics of pneumatic musclerdquoAnnals of Biomedical Engineering vol 31 no 3 pp 310ndash3172003
[14] C S Tseng B S Chen and H J Uang ldquoFuzzy tracking controldesign for nonlinear dynamic systems via T-S fuzzy modelrdquoIEEE Transactions on Fuzzy Systems vol 9 no 3 pp 381ndash3922001
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
The Scientific World Journal 5
1
2
3
4
5
6
78910
11
12
1314
15
16
ADDA
Figure 1 The experimental setup
1b
1205911
2b
1205912 m2
r1
1205792
2a
1a
m1
r1
1205791
l1
l2
p1b = p10 minus Δp1
p1a = p10 + Δp1
p2a=p 20
+ Δp2
p 2b=p 20
minus Δp2
Figure 2 Operation principle of the leg rehabilitation machine driven by PAMs
6 The Scientific World Journal
Table 1 Component specifications
Number Component Specifications1 2 3 4 PAM Festo MAS-20-150N7 8 9 10 Pressure proportional valve Mac PPC5C5 6 Rotary potentiometer Keen Engineering KRT205011 12 13 14 Pressure transducer Jihsense SN 911166 0ndash10 kgfcm2
16 IBM-compatible PC Pentium 18GHzDAC ADC Automation AIO3321
where
1206011119886
= 1198651119886(1199011119886) minus 1198701119886(1199011119886) 11990311205791minus 1198611119886(1199011119886) 1199031
1205791
(29a)
1206011119887
= 1198651119887(1199011119887) minus 1198701119887(1199011119887) 11990311205791minus 1198611119887(1199011119887) 1199031
1205791
(29b)
1206012119886
= 1198652119886(1199012119886) minus 1198702119886(1199012119886) 11990321205792minus 1198612119886(1199012119886) 1199032
1205792
(29c)
1206012119887
= 1198652119887(1199012119887) minus 1198702119887(1199012119887) 11990321205792minus 1198612119887(1199012119887) 1199032
1205792 (29d)
where the spring coefficient119870(119901) and the damping coefficient119861(119901) are given by Reynolds et al [13]
The desired input pressures P119886= [1199011119886
1199012119886]119879 and P
119887=
[1199011119887
1199012119887]119879 for each PAM are generated by the following
equation
P119886(119905) = P
0+ ΔP (119905) P
119887(119905) = P
0minus ΔP (119905) (30)
where P0
= [11990110
11990120]119879 is a nominal constant input PAM
pressure and ΔP(119905) = [Δ1199011
Δ1199012]119879 is the control pressure
input with an arbitrary function of time Because the pressureinputΔP(119905) = [Δ119901
1Δ1199012]119879 and output 120579 = [120579
11205792]119879 the sys-
tem can bewritten as a two-input two-output (TITO) controlsystem The control signal u = [119906
11199062]119879 is proportional to
ΔP based on the pressure proportional valversquos characteristicsThat is ΔP can be used instead of u as a control input
4 Dynamic Model of a Two-Joint LegRehabilitation Machine Driven by PAMs
Figure 2 shows a two-joint leg rehabilitation machine drivenby PAMs and the dynamic equation is given as follows [14]
119872(120579) 120579 + 119862 (120579 120579) 120579 + 119866 (120579) = 120591 (31)
where
119872(120579) = [(1198981+ 1198982) 1198972
119898211989711198972(11990411199042+ 11988811198882)
119898211989711198972(11990411199042+ 11988811198882) 119898
21198972
2
]
119862 (120579 120579) = [0 minus119898
211989711198972(11988811199042minus 11990411198882) 1205792
minus119898211989711198972(11988811199042minus 11990411198882) 1205791
0]
119866 (120579) = [(1198981+ 1198982) 11989711198921199041
minus119898211989721198921199042
]
(32)
and119872(120579) is the moment of inertia 119862(120579 120579) includes Coriolisand centripetal force and 119866(120579) is the gravitational force
Notation 1199041= sin(120579
1) 1199042= sin(120579
2) 1198881= cos(120579
1) and 119888
2=
cos(1205792) Let 119909
1= 1205791 1199092= 1205791 1199093= 1205792 and 119909
4= 1205792 then (31)
can be written as the following state-space form [14]
1= 1199092
2= 1198911(119909) + 119892
11(119909) 1205911+ 119892121205912
3= 1199094
4= 1198912(119909) + 119892
21(119909) + 119892
221205912
(33)
where1198911(119909)
=
(11990411198882minus 11988811199042) [119898211989711198972(11990411199042+ 11988811198882) 1199092
2minus 11989821198972
21199092
4]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
+[(1198981minus 1198982) 11989721198921199041minus 119898211989721198921199042(11990411199042+ 11988811198882)]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
1198912(119909)
=
(11990411198882minus 11988811199042) [minus (119898
1+ 1198982) 1198972
11199092
2+ 119898211989711198972(11990411199042+ 11988811198882) 1199092
4]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
+[minus (119898
1+ 1198982) 11989711198921199041(11990411199042+ 11988811198882) + (119898
1+ 1198982) 11989711198921199042]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989211(119909) =
11989821198972
2
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989212(119909) =
minus119898211989711198972(11990411199042+ 11988811198882)
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989221(119909) =
minus119898211989711198972(11990411199042+ 11988811198882)
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989222(119909) =
(1198981+ 1198982) 1198972
1
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
(34)
5 Experimental Studies
Figure 3 shows the leg rehabilitation machine using an actualhuman loading with a 65 kg weight The automatic device
The Scientific World Journal 7
Figure 3 The two-joint leg rehabilitation device with actual human loading
can help patients to recover lower limb motion function bymeans of continuous passive motion such as a sinusoidalwave command an irregular curve command and an end-effect tracking command The experiments include both theproposed approach and PDC for comparison in order toevaluate efficacy and control performance The controllerswere implemented on an Intel Pentium 18GHz PC with asampling time of 5ms and the entire control software wascoded in C++
This study attempts to use as few rules as possible inorder tominimize design effort and complexityTheT-S fuzzymodel of the system is thus given the following four-rulefuzzy model
1198771
IF 1199091is about (1205874) and 119909
3is about (1205874)
THEN = 11986011199091+ 1198611119906
1198772
IF 1199091is about (1205874) and 119909
3is about (minus1205874)
THEN = 11986021199091+ 1198612119906
1198773
IF 1199091is about (minus1205874) and 119909
3is about (1205874)
THEN = 11986031199091+ 1198613119906
1198774
IF 1199091is about (minus1205874) and 119909
3is about (minus1205874)
THEN = 11986041199091+ 1198614119906
(35)
where
1198601=
[[[
[
0 1 0 0
44082 00059 06742 minus00002
0 0 0 1
minus14572 00002 minus39722 00002
]]]
]
1198602=
[[[
[
0 1 0 0
46734 00049 0532 00001
0 0 0 1
minus12145 minus00001 minus3563 00001
]]]
]
1198603=
[[[
[
0 1 0 0
54782 00021 06876 minus00002
0 0 0 1
14525 00002 46723 00002
]]]
]
1198604=
[[[[
[
0 1 0 0
49821 minus00044 06543 00002
0 0 0 1
minus11034 minus00002 35631 00003
]]]]
]
1198611= 1198614=
[[[[
[
0 0
14811 minus2849
0 0
minus2849 237417
]]]]
]
1198612= 1198613=
[[[[
[
0 0
11965 10297
0 0
10297 191501
]]]]
]
119875 =
[[[[
[
162602 minus06640 10734 minus00095
minus06640 03007 minus03455 02871
10734 minus03455 04730 minus03458
minus00095 02871 minus03458 03658
]]]]
]
1198701= [
minus04509 01745 minus07182 00527
minus02807 00034 minus00471 02639]
1198702= [
minus04619 01238 minus07812 00351
minus02827 00068 minus00851 01401]
1198703= [
minus05902 02132 minus08910 00531
minus04107 00117 minus00730 02989]
1198704= [
minus04891 02109 minus06734 004145
minus03180 00085 minus00631 03893]
(36)
which guarantee the stability condition (17) MATLAB Tool-box is used to obtain parameters as 119896
1= 00251 and
8 The Scientific World Journal
Table 2 Peak-peak error and phase lag for Figure 5
The proposed approach PDCPeak-peak error Phase lag Peak-peak error Phase lag
1205791
1205792
1205791
1205792
1205791
1205792
1205791
1205792
07 035 49∘ 45∘ 13 065 162∘ 108∘
20
10
0
minus10
minus20
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
ReferenceActual
15
0
minus15
1205792
(a) The proposed approach
20
10
0
minus10
minus20
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
ReferenceActual
15
0
minus15
1205792
(b) PDC
Figure 4 Sinusoidal wave response for knee and ankle joints
1198962= 001539 For comparison with the proposed controller
the PDC feedback gains are designed to be
1= [
minus03293 01023 minus02538 002317
minus01783 00021 minus002672 04732]
1198702= [
minus04278 01845 minus02451 00731
minus02185 00026 minus00752 00923]
1198703= [
minus06013 03461 minus05482 00231
minus04421 00093 minus00651 03529]
1198704= [
minus03756 03150 minus05391 00421
minus04250 00023 minus00531 03597]
(37)
51 Sinusoidal Wave Response Continuous reciprocation isrequired in order to foster the recovery of extremity functionThe sinusoidal wave responses of the proposed approach andPDC for both knee and ankle joints are shown in Figure 4It is evident that angle trajectories of the proposed approachare close to the command Figure 5 shows that the proposedapproach exhibits less tracking errors than does PDC Thepeak-peak error and phase lag are listed in Table 2 Because
of the interaction of the two joints PDC has significant angleerrors for 120579
1 which will degrade the rehabilitation effect
However supervisory control can overcome the couplingeffect of the two joints to achieve excellent rehabilitationfunction for patients
52 Irregular Curve Response In practical applications itcould be expected that the reference command will changewith different input frequencies The desired trajectories forboth knee and ankle joints are
1205791= 20 lowast 033 (sin (2120587119891
1119905) + sin (2120587119891
2119905) + sin (2120587119891
3119905))
1205792= 15 lowast 033 (sin (2120587119891
1119905) + sin (2120587119891
2119905) + sin (2120587119891
3119905))
(38)
with 1198911= 005Hz 119891
2= 01Hz and 119891
3= 0066Hz
Figure 6 shows the tracking responses of irregular curvesobtained using both the proposed approach and PDC Track-ing errors for the knee and ankle joints are shown in Figure 7Clearly the angle error of the proposed approach is averagemaintained within 2∘ However the proposed approach iscapable of adapting to different frequencies
The Scientific World Journal 9
14
12
10
8
6
4
2
0
minus2
minus41009080706050403020100
Time (s)
Ang
le (d
eg)
1205791 error1205792 error
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 5 Angle tracking errors for both knee and ankle
20
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
120579215
10
5
0
minus5
minus10
minus15
Actual
(a) The proposed approach
20
10
0
minus10
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
1205792
Actual
(b) PDC
Figure 6 Irregular curve response for both knee and ankle joints
53 Elliptic Response The desired end-effect or trajectory isgiven by
119909119889(119905) = 0614 minus 0015 sdot cos (02120587 sdot 119905 minus 120587)
119910119889(119905) = minus01 sdot sin (02120587 sdot 119905 minus 2120587)
(39)
where 0 le 119905 le 20 seconds
The end-effect tracking responses in the 119909 119910 coordinatefor both the proposed approach and PDC are shown inFigure 8 and the end-effect position tracking errors aredisplayed in Figure 9 It is evident that tracking behavior ofthe proposed approach is better than that of the PDC Ascan be seen the tracking errors of the proposed approach are
10 The Scientific World Journal
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 7 Angle tracking errors for both the proposed approach and PDC
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(a) The proposed approach
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(b) PDC
Figure 8 Elliptic response of the proposed approach and PDC
within 003m On the other hand angle tracking errors ofknee and ankle joints are shown in Figure 10
Moreover it is difficult to enhance end-effector trackingperformance using the PDC algorithm because the PDCcannot overcome the nonlinearity of PAMs and the structuralinteraction However the proposed approach overcomessuccessfully the coupling effect and parameter uncertaintiesof the system As seen in the experimental results theproposed approach can attain excellent end-effector trackingperformance in rehabilitation function
6 Conclusions
In this study a novel composite fuzzy control is proposed andapplied in the two-joint leg rehabilitation device driven byPAMsThe proposed controller is not only capable of decom-posing nonlinear systems into a set of linear subsystems but isalso capable of simplifying a complex nonlinear system usinglinear control techniques with the control gains determinedusingMATLABrsquos LMI Toolbox based on the Lyapunov stabil-ity theoremMoreover the supervisory control can overcome
The Scientific World Journal 11
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(a) The proposed approach
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(b) PDC
Figure 9 The end-effect position tracking errors
14
12
10
8
6
4
2
0
minus2
minus40 2 4 6 8 10 12 14 16 18 20
Ang
le (d
eg)
Time (s)
1205791 error1205792 error
(a) The proposed approach
0 2 4 6 8 10 12 14 16 18 20
Time (s)
15
10
5
0
minus5
minus10
1205791 error1205792 error
Ang
le (d
eg)
(b) PDC
Figure 10 Angle tracking errors of knee and ankle joints
the coupling effect for a leg rehabilitation machine Experi-mental results show that the system response of the proposedapproach was in good agreement with that of the refer-ence input and guarantee robustness to system parameteruncertainties
Conflict of Interests
The author declares that there is no conflict of interestsregarding the publication of this paper
References
[1] D G Caldwell G A Medrano-Cerda and M GoodwinldquoControl of pneumatic muscle actuatorsrdquo IEEE Control SystemsMagazine vol 15 no 1 pp 40ndash48 1995
[2] C P Chou and B Hannaford ldquoStatic and dynamic characteris-tics ofMcKibben pneumatic artificialmusclesrdquo inProceedings ofthe IEEE International Conference on Robotics and Automationpp 281ndash286 May 1994
[3] T Noritsugu and T Tanaka ldquoApplication of rubber artificialmuscle manipulator as a rehabilitation robotrdquo IEEEASMETransactions on Mechatronics vol 2 no 4 pp 259ndash267 1997
12 The Scientific World Journal
[4] J H Lilly and L Yang ldquoSliding mode tracking for pneumaticmuscle actuators in opposing pair configurationrdquo IEEE Trans-actions on Control Systems Technology vol 4 pp 550ndash558 2005
[5] K K Ahn and H P H Anh ldquoDesign and implementation of anadaptive recurrent neural networks (ARNN) controller of thepneumatic artificial muscle (PAM) manipulatorrdquoMechatronicsvol 19 no 6 pp 816ndash828 2009
[6] X Shen ldquoNonlinearmodel-based control of pneumatic artificialmuscle servo systemsrdquo Control Engineering Practice vol 18 no3 pp 311ndash317 2010
[7] L A Zadeh ldquoFuzzy setsrdquo Information and Control vol 8 no 3pp 338ndash353 1965
[8] T Takagi and M Sugeno ldquoFuzzy identification of systems andits applications to modeling and controlrdquo IEEE Transactions onSystems Man and Cybernetics vol 15 no 1 pp 116ndash132 1985
[9] K K Ahn and H P H Anh ldquoInverse double NARX fuzzymodeling for system identificationrdquo IEEEASME Transactionson Mechatronics vol 15 no 1 pp 136ndash148 2010
[10] L Seddiki K Guelton and J Zaytoon ldquoConcept and Takagi-Sugeno descriptor tracking controller design of a closedmuscu-lar chain lower-limb rehabilitation devicerdquo IET Control Theoryand Applications vol 4 no 8 pp 1407ndash1420 2010
[11] K Tanaka T Ikeda and H O Wang ldquoFuzzy regulators andfuzzy observers relaxed stability conditions and LMI-baseddesignsrdquo IEEE Transactions on Fuzzy Systems vol 6 no 2 pp250ndash265 1998
[12] M K Chang J J Liou andM L Chen ldquoT-S fuzzymodel-basedtracking control of a one-dimensional manipulator actuated bypneumatic artificial musclesrdquo Control Engineering Practice vol19 no 12 pp 1442ndash1449 2011
[13] D B Reynolds D W Repperger C A Phillips and G BandryldquoModeling the dynamic characteristics of pneumatic musclerdquoAnnals of Biomedical Engineering vol 31 no 3 pp 310ndash3172003
[14] C S Tseng B S Chen and H J Uang ldquoFuzzy tracking controldesign for nonlinear dynamic systems via T-S fuzzy modelrdquoIEEE Transactions on Fuzzy Systems vol 9 no 3 pp 381ndash3922001
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
6 The Scientific World Journal
Table 1 Component specifications
Number Component Specifications1 2 3 4 PAM Festo MAS-20-150N7 8 9 10 Pressure proportional valve Mac PPC5C5 6 Rotary potentiometer Keen Engineering KRT205011 12 13 14 Pressure transducer Jihsense SN 911166 0ndash10 kgfcm2
16 IBM-compatible PC Pentium 18GHzDAC ADC Automation AIO3321
where
1206011119886
= 1198651119886(1199011119886) minus 1198701119886(1199011119886) 11990311205791minus 1198611119886(1199011119886) 1199031
1205791
(29a)
1206011119887
= 1198651119887(1199011119887) minus 1198701119887(1199011119887) 11990311205791minus 1198611119887(1199011119887) 1199031
1205791
(29b)
1206012119886
= 1198652119886(1199012119886) minus 1198702119886(1199012119886) 11990321205792minus 1198612119886(1199012119886) 1199032
1205792
(29c)
1206012119887
= 1198652119887(1199012119887) minus 1198702119887(1199012119887) 11990321205792minus 1198612119887(1199012119887) 1199032
1205792 (29d)
where the spring coefficient119870(119901) and the damping coefficient119861(119901) are given by Reynolds et al [13]
The desired input pressures P119886= [1199011119886
1199012119886]119879 and P
119887=
[1199011119887
1199012119887]119879 for each PAM are generated by the following
equation
P119886(119905) = P
0+ ΔP (119905) P
119887(119905) = P
0minus ΔP (119905) (30)
where P0
= [11990110
11990120]119879 is a nominal constant input PAM
pressure and ΔP(119905) = [Δ1199011
Δ1199012]119879 is the control pressure
input with an arbitrary function of time Because the pressureinputΔP(119905) = [Δ119901
1Δ1199012]119879 and output 120579 = [120579
11205792]119879 the sys-
tem can bewritten as a two-input two-output (TITO) controlsystem The control signal u = [119906
11199062]119879 is proportional to
ΔP based on the pressure proportional valversquos characteristicsThat is ΔP can be used instead of u as a control input
4 Dynamic Model of a Two-Joint LegRehabilitation Machine Driven by PAMs
Figure 2 shows a two-joint leg rehabilitation machine drivenby PAMs and the dynamic equation is given as follows [14]
119872(120579) 120579 + 119862 (120579 120579) 120579 + 119866 (120579) = 120591 (31)
where
119872(120579) = [(1198981+ 1198982) 1198972
119898211989711198972(11990411199042+ 11988811198882)
119898211989711198972(11990411199042+ 11988811198882) 119898
21198972
2
]
119862 (120579 120579) = [0 minus119898
211989711198972(11988811199042minus 11990411198882) 1205792
minus119898211989711198972(11988811199042minus 11990411198882) 1205791
0]
119866 (120579) = [(1198981+ 1198982) 11989711198921199041
minus119898211989721198921199042
]
(32)
and119872(120579) is the moment of inertia 119862(120579 120579) includes Coriolisand centripetal force and 119866(120579) is the gravitational force
Notation 1199041= sin(120579
1) 1199042= sin(120579
2) 1198881= cos(120579
1) and 119888
2=
cos(1205792) Let 119909
1= 1205791 1199092= 1205791 1199093= 1205792 and 119909
4= 1205792 then (31)
can be written as the following state-space form [14]
1= 1199092
2= 1198911(119909) + 119892
11(119909) 1205911+ 119892121205912
3= 1199094
4= 1198912(119909) + 119892
21(119909) + 119892
221205912
(33)
where1198911(119909)
=
(11990411198882minus 11988811199042) [119898211989711198972(11990411199042+ 11988811198882) 1199092
2minus 11989821198972
21199092
4]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
+[(1198981minus 1198982) 11989721198921199041minus 119898211989721198921199042(11990411199042+ 11988811198882)]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
1198912(119909)
=
(11990411198882minus 11988811199042) [minus (119898
1+ 1198982) 1198972
11199092
2+ 119898211989711198972(11990411199042+ 11988811198882) 1199092
4]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
+[minus (119898
1+ 1198982) 11989711198921199041(11990411199042+ 11988811198882) + (119898
1+ 1198982) 11989711198921199042]
11989711198972[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989211(119909) =
11989821198972
2
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989212(119909) =
minus119898211989711198972(11990411199042+ 11988811198882)
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989221(119909) =
minus119898211989711198972(11990411199042+ 11988811198882)
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
11989222(119909) =
(1198981+ 1198982) 1198972
1
11989821198972
11198972
2[(1198981+ 1198982) minus 1198982(11990411199042+ 11988811198882)2
]
(34)
5 Experimental Studies
Figure 3 shows the leg rehabilitation machine using an actualhuman loading with a 65 kg weight The automatic device
The Scientific World Journal 7
Figure 3 The two-joint leg rehabilitation device with actual human loading
can help patients to recover lower limb motion function bymeans of continuous passive motion such as a sinusoidalwave command an irregular curve command and an end-effect tracking command The experiments include both theproposed approach and PDC for comparison in order toevaluate efficacy and control performance The controllerswere implemented on an Intel Pentium 18GHz PC with asampling time of 5ms and the entire control software wascoded in C++
This study attempts to use as few rules as possible inorder tominimize design effort and complexityTheT-S fuzzymodel of the system is thus given the following four-rulefuzzy model
1198771
IF 1199091is about (1205874) and 119909
3is about (1205874)
THEN = 11986011199091+ 1198611119906
1198772
IF 1199091is about (1205874) and 119909
3is about (minus1205874)
THEN = 11986021199091+ 1198612119906
1198773
IF 1199091is about (minus1205874) and 119909
3is about (1205874)
THEN = 11986031199091+ 1198613119906
1198774
IF 1199091is about (minus1205874) and 119909
3is about (minus1205874)
THEN = 11986041199091+ 1198614119906
(35)
where
1198601=
[[[
[
0 1 0 0
44082 00059 06742 minus00002
0 0 0 1
minus14572 00002 minus39722 00002
]]]
]
1198602=
[[[
[
0 1 0 0
46734 00049 0532 00001
0 0 0 1
minus12145 minus00001 minus3563 00001
]]]
]
1198603=
[[[
[
0 1 0 0
54782 00021 06876 minus00002
0 0 0 1
14525 00002 46723 00002
]]]
]
1198604=
[[[[
[
0 1 0 0
49821 minus00044 06543 00002
0 0 0 1
minus11034 minus00002 35631 00003
]]]]
]
1198611= 1198614=
[[[[
[
0 0
14811 minus2849
0 0
minus2849 237417
]]]]
]
1198612= 1198613=
[[[[
[
0 0
11965 10297
0 0
10297 191501
]]]]
]
119875 =
[[[[
[
162602 minus06640 10734 minus00095
minus06640 03007 minus03455 02871
10734 minus03455 04730 minus03458
minus00095 02871 minus03458 03658
]]]]
]
1198701= [
minus04509 01745 minus07182 00527
minus02807 00034 minus00471 02639]
1198702= [
minus04619 01238 minus07812 00351
minus02827 00068 minus00851 01401]
1198703= [
minus05902 02132 minus08910 00531
minus04107 00117 minus00730 02989]
1198704= [
minus04891 02109 minus06734 004145
minus03180 00085 minus00631 03893]
(36)
which guarantee the stability condition (17) MATLAB Tool-box is used to obtain parameters as 119896
1= 00251 and
8 The Scientific World Journal
Table 2 Peak-peak error and phase lag for Figure 5
The proposed approach PDCPeak-peak error Phase lag Peak-peak error Phase lag
1205791
1205792
1205791
1205792
1205791
1205792
1205791
1205792
07 035 49∘ 45∘ 13 065 162∘ 108∘
20
10
0
minus10
minus20
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
ReferenceActual
15
0
minus15
1205792
(a) The proposed approach
20
10
0
minus10
minus20
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
ReferenceActual
15
0
minus15
1205792
(b) PDC
Figure 4 Sinusoidal wave response for knee and ankle joints
1198962= 001539 For comparison with the proposed controller
the PDC feedback gains are designed to be
1= [
minus03293 01023 minus02538 002317
minus01783 00021 minus002672 04732]
1198702= [
minus04278 01845 minus02451 00731
minus02185 00026 minus00752 00923]
1198703= [
minus06013 03461 minus05482 00231
minus04421 00093 minus00651 03529]
1198704= [
minus03756 03150 minus05391 00421
minus04250 00023 minus00531 03597]
(37)
51 Sinusoidal Wave Response Continuous reciprocation isrequired in order to foster the recovery of extremity functionThe sinusoidal wave responses of the proposed approach andPDC for both knee and ankle joints are shown in Figure 4It is evident that angle trajectories of the proposed approachare close to the command Figure 5 shows that the proposedapproach exhibits less tracking errors than does PDC Thepeak-peak error and phase lag are listed in Table 2 Because
of the interaction of the two joints PDC has significant angleerrors for 120579
1 which will degrade the rehabilitation effect
However supervisory control can overcome the couplingeffect of the two joints to achieve excellent rehabilitationfunction for patients
52 Irregular Curve Response In practical applications itcould be expected that the reference command will changewith different input frequencies The desired trajectories forboth knee and ankle joints are
1205791= 20 lowast 033 (sin (2120587119891
1119905) + sin (2120587119891
2119905) + sin (2120587119891
3119905))
1205792= 15 lowast 033 (sin (2120587119891
1119905) + sin (2120587119891
2119905) + sin (2120587119891
3119905))
(38)
with 1198911= 005Hz 119891
2= 01Hz and 119891
3= 0066Hz
Figure 6 shows the tracking responses of irregular curvesobtained using both the proposed approach and PDC Track-ing errors for the knee and ankle joints are shown in Figure 7Clearly the angle error of the proposed approach is averagemaintained within 2∘ However the proposed approach iscapable of adapting to different frequencies
The Scientific World Journal 9
14
12
10
8
6
4
2
0
minus2
minus41009080706050403020100
Time (s)
Ang
le (d
eg)
1205791 error1205792 error
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 5 Angle tracking errors for both knee and ankle
20
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
120579215
10
5
0
minus5
minus10
minus15
Actual
(a) The proposed approach
20
10
0
minus10
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
1205792
Actual
(b) PDC
Figure 6 Irregular curve response for both knee and ankle joints
53 Elliptic Response The desired end-effect or trajectory isgiven by
119909119889(119905) = 0614 minus 0015 sdot cos (02120587 sdot 119905 minus 120587)
119910119889(119905) = minus01 sdot sin (02120587 sdot 119905 minus 2120587)
(39)
where 0 le 119905 le 20 seconds
The end-effect tracking responses in the 119909 119910 coordinatefor both the proposed approach and PDC are shown inFigure 8 and the end-effect position tracking errors aredisplayed in Figure 9 It is evident that tracking behavior ofthe proposed approach is better than that of the PDC Ascan be seen the tracking errors of the proposed approach are
10 The Scientific World Journal
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 7 Angle tracking errors for both the proposed approach and PDC
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(a) The proposed approach
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(b) PDC
Figure 8 Elliptic response of the proposed approach and PDC
within 003m On the other hand angle tracking errors ofknee and ankle joints are shown in Figure 10
Moreover it is difficult to enhance end-effector trackingperformance using the PDC algorithm because the PDCcannot overcome the nonlinearity of PAMs and the structuralinteraction However the proposed approach overcomessuccessfully the coupling effect and parameter uncertaintiesof the system As seen in the experimental results theproposed approach can attain excellent end-effector trackingperformance in rehabilitation function
6 Conclusions
In this study a novel composite fuzzy control is proposed andapplied in the two-joint leg rehabilitation device driven byPAMsThe proposed controller is not only capable of decom-posing nonlinear systems into a set of linear subsystems but isalso capable of simplifying a complex nonlinear system usinglinear control techniques with the control gains determinedusingMATLABrsquos LMI Toolbox based on the Lyapunov stabil-ity theoremMoreover the supervisory control can overcome
The Scientific World Journal 11
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(a) The proposed approach
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(b) PDC
Figure 9 The end-effect position tracking errors
14
12
10
8
6
4
2
0
minus2
minus40 2 4 6 8 10 12 14 16 18 20
Ang
le (d
eg)
Time (s)
1205791 error1205792 error
(a) The proposed approach
0 2 4 6 8 10 12 14 16 18 20
Time (s)
15
10
5
0
minus5
minus10
1205791 error1205792 error
Ang
le (d
eg)
(b) PDC
Figure 10 Angle tracking errors of knee and ankle joints
the coupling effect for a leg rehabilitation machine Experi-mental results show that the system response of the proposedapproach was in good agreement with that of the refer-ence input and guarantee robustness to system parameteruncertainties
Conflict of Interests
The author declares that there is no conflict of interestsregarding the publication of this paper
References
[1] D G Caldwell G A Medrano-Cerda and M GoodwinldquoControl of pneumatic muscle actuatorsrdquo IEEE Control SystemsMagazine vol 15 no 1 pp 40ndash48 1995
[2] C P Chou and B Hannaford ldquoStatic and dynamic characteris-tics ofMcKibben pneumatic artificialmusclesrdquo inProceedings ofthe IEEE International Conference on Robotics and Automationpp 281ndash286 May 1994
[3] T Noritsugu and T Tanaka ldquoApplication of rubber artificialmuscle manipulator as a rehabilitation robotrdquo IEEEASMETransactions on Mechatronics vol 2 no 4 pp 259ndash267 1997
12 The Scientific World Journal
[4] J H Lilly and L Yang ldquoSliding mode tracking for pneumaticmuscle actuators in opposing pair configurationrdquo IEEE Trans-actions on Control Systems Technology vol 4 pp 550ndash558 2005
[5] K K Ahn and H P H Anh ldquoDesign and implementation of anadaptive recurrent neural networks (ARNN) controller of thepneumatic artificial muscle (PAM) manipulatorrdquoMechatronicsvol 19 no 6 pp 816ndash828 2009
[6] X Shen ldquoNonlinearmodel-based control of pneumatic artificialmuscle servo systemsrdquo Control Engineering Practice vol 18 no3 pp 311ndash317 2010
[7] L A Zadeh ldquoFuzzy setsrdquo Information and Control vol 8 no 3pp 338ndash353 1965
[8] T Takagi and M Sugeno ldquoFuzzy identification of systems andits applications to modeling and controlrdquo IEEE Transactions onSystems Man and Cybernetics vol 15 no 1 pp 116ndash132 1985
[9] K K Ahn and H P H Anh ldquoInverse double NARX fuzzymodeling for system identificationrdquo IEEEASME Transactionson Mechatronics vol 15 no 1 pp 136ndash148 2010
[10] L Seddiki K Guelton and J Zaytoon ldquoConcept and Takagi-Sugeno descriptor tracking controller design of a closedmuscu-lar chain lower-limb rehabilitation devicerdquo IET Control Theoryand Applications vol 4 no 8 pp 1407ndash1420 2010
[11] K Tanaka T Ikeda and H O Wang ldquoFuzzy regulators andfuzzy observers relaxed stability conditions and LMI-baseddesignsrdquo IEEE Transactions on Fuzzy Systems vol 6 no 2 pp250ndash265 1998
[12] M K Chang J J Liou andM L Chen ldquoT-S fuzzymodel-basedtracking control of a one-dimensional manipulator actuated bypneumatic artificial musclesrdquo Control Engineering Practice vol19 no 12 pp 1442ndash1449 2011
[13] D B Reynolds D W Repperger C A Phillips and G BandryldquoModeling the dynamic characteristics of pneumatic musclerdquoAnnals of Biomedical Engineering vol 31 no 3 pp 310ndash3172003
[14] C S Tseng B S Chen and H J Uang ldquoFuzzy tracking controldesign for nonlinear dynamic systems via T-S fuzzy modelrdquoIEEE Transactions on Fuzzy Systems vol 9 no 3 pp 381ndash3922001
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
The Scientific World Journal 7
Figure 3 The two-joint leg rehabilitation device with actual human loading
can help patients to recover lower limb motion function bymeans of continuous passive motion such as a sinusoidalwave command an irregular curve command and an end-effect tracking command The experiments include both theproposed approach and PDC for comparison in order toevaluate efficacy and control performance The controllerswere implemented on an Intel Pentium 18GHz PC with asampling time of 5ms and the entire control software wascoded in C++
This study attempts to use as few rules as possible inorder tominimize design effort and complexityTheT-S fuzzymodel of the system is thus given the following four-rulefuzzy model
1198771
IF 1199091is about (1205874) and 119909
3is about (1205874)
THEN = 11986011199091+ 1198611119906
1198772
IF 1199091is about (1205874) and 119909
3is about (minus1205874)
THEN = 11986021199091+ 1198612119906
1198773
IF 1199091is about (minus1205874) and 119909
3is about (1205874)
THEN = 11986031199091+ 1198613119906
1198774
IF 1199091is about (minus1205874) and 119909
3is about (minus1205874)
THEN = 11986041199091+ 1198614119906
(35)
where
1198601=
[[[
[
0 1 0 0
44082 00059 06742 minus00002
0 0 0 1
minus14572 00002 minus39722 00002
]]]
]
1198602=
[[[
[
0 1 0 0
46734 00049 0532 00001
0 0 0 1
minus12145 minus00001 minus3563 00001
]]]
]
1198603=
[[[
[
0 1 0 0
54782 00021 06876 minus00002
0 0 0 1
14525 00002 46723 00002
]]]
]
1198604=
[[[[
[
0 1 0 0
49821 minus00044 06543 00002
0 0 0 1
minus11034 minus00002 35631 00003
]]]]
]
1198611= 1198614=
[[[[
[
0 0
14811 minus2849
0 0
minus2849 237417
]]]]
]
1198612= 1198613=
[[[[
[
0 0
11965 10297
0 0
10297 191501
]]]]
]
119875 =
[[[[
[
162602 minus06640 10734 minus00095
minus06640 03007 minus03455 02871
10734 minus03455 04730 minus03458
minus00095 02871 minus03458 03658
]]]]
]
1198701= [
minus04509 01745 minus07182 00527
minus02807 00034 minus00471 02639]
1198702= [
minus04619 01238 minus07812 00351
minus02827 00068 minus00851 01401]
1198703= [
minus05902 02132 minus08910 00531
minus04107 00117 minus00730 02989]
1198704= [
minus04891 02109 minus06734 004145
minus03180 00085 minus00631 03893]
(36)
which guarantee the stability condition (17) MATLAB Tool-box is used to obtain parameters as 119896
1= 00251 and
8 The Scientific World Journal
Table 2 Peak-peak error and phase lag for Figure 5
The proposed approach PDCPeak-peak error Phase lag Peak-peak error Phase lag
1205791
1205792
1205791
1205792
1205791
1205792
1205791
1205792
07 035 49∘ 45∘ 13 065 162∘ 108∘
20
10
0
minus10
minus20
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
ReferenceActual
15
0
minus15
1205792
(a) The proposed approach
20
10
0
minus10
minus20
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
ReferenceActual
15
0
minus15
1205792
(b) PDC
Figure 4 Sinusoidal wave response for knee and ankle joints
1198962= 001539 For comparison with the proposed controller
the PDC feedback gains are designed to be
1= [
minus03293 01023 minus02538 002317
minus01783 00021 minus002672 04732]
1198702= [
minus04278 01845 minus02451 00731
minus02185 00026 minus00752 00923]
1198703= [
minus06013 03461 minus05482 00231
minus04421 00093 minus00651 03529]
1198704= [
minus03756 03150 minus05391 00421
minus04250 00023 minus00531 03597]
(37)
51 Sinusoidal Wave Response Continuous reciprocation isrequired in order to foster the recovery of extremity functionThe sinusoidal wave responses of the proposed approach andPDC for both knee and ankle joints are shown in Figure 4It is evident that angle trajectories of the proposed approachare close to the command Figure 5 shows that the proposedapproach exhibits less tracking errors than does PDC Thepeak-peak error and phase lag are listed in Table 2 Because
of the interaction of the two joints PDC has significant angleerrors for 120579
1 which will degrade the rehabilitation effect
However supervisory control can overcome the couplingeffect of the two joints to achieve excellent rehabilitationfunction for patients
52 Irregular Curve Response In practical applications itcould be expected that the reference command will changewith different input frequencies The desired trajectories forboth knee and ankle joints are
1205791= 20 lowast 033 (sin (2120587119891
1119905) + sin (2120587119891
2119905) + sin (2120587119891
3119905))
1205792= 15 lowast 033 (sin (2120587119891
1119905) + sin (2120587119891
2119905) + sin (2120587119891
3119905))
(38)
with 1198911= 005Hz 119891
2= 01Hz and 119891
3= 0066Hz
Figure 6 shows the tracking responses of irregular curvesobtained using both the proposed approach and PDC Track-ing errors for the knee and ankle joints are shown in Figure 7Clearly the angle error of the proposed approach is averagemaintained within 2∘ However the proposed approach iscapable of adapting to different frequencies
The Scientific World Journal 9
14
12
10
8
6
4
2
0
minus2
minus41009080706050403020100
Time (s)
Ang
le (d
eg)
1205791 error1205792 error
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 5 Angle tracking errors for both knee and ankle
20
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
120579215
10
5
0
minus5
minus10
minus15
Actual
(a) The proposed approach
20
10
0
minus10
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
1205792
Actual
(b) PDC
Figure 6 Irregular curve response for both knee and ankle joints
53 Elliptic Response The desired end-effect or trajectory isgiven by
119909119889(119905) = 0614 minus 0015 sdot cos (02120587 sdot 119905 minus 120587)
119910119889(119905) = minus01 sdot sin (02120587 sdot 119905 minus 2120587)
(39)
where 0 le 119905 le 20 seconds
The end-effect tracking responses in the 119909 119910 coordinatefor both the proposed approach and PDC are shown inFigure 8 and the end-effect position tracking errors aredisplayed in Figure 9 It is evident that tracking behavior ofthe proposed approach is better than that of the PDC Ascan be seen the tracking errors of the proposed approach are
10 The Scientific World Journal
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 7 Angle tracking errors for both the proposed approach and PDC
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(a) The proposed approach
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(b) PDC
Figure 8 Elliptic response of the proposed approach and PDC
within 003m On the other hand angle tracking errors ofknee and ankle joints are shown in Figure 10
Moreover it is difficult to enhance end-effector trackingperformance using the PDC algorithm because the PDCcannot overcome the nonlinearity of PAMs and the structuralinteraction However the proposed approach overcomessuccessfully the coupling effect and parameter uncertaintiesof the system As seen in the experimental results theproposed approach can attain excellent end-effector trackingperformance in rehabilitation function
6 Conclusions
In this study a novel composite fuzzy control is proposed andapplied in the two-joint leg rehabilitation device driven byPAMsThe proposed controller is not only capable of decom-posing nonlinear systems into a set of linear subsystems but isalso capable of simplifying a complex nonlinear system usinglinear control techniques with the control gains determinedusingMATLABrsquos LMI Toolbox based on the Lyapunov stabil-ity theoremMoreover the supervisory control can overcome
The Scientific World Journal 11
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(a) The proposed approach
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(b) PDC
Figure 9 The end-effect position tracking errors
14
12
10
8
6
4
2
0
minus2
minus40 2 4 6 8 10 12 14 16 18 20
Ang
le (d
eg)
Time (s)
1205791 error1205792 error
(a) The proposed approach
0 2 4 6 8 10 12 14 16 18 20
Time (s)
15
10
5
0
minus5
minus10
1205791 error1205792 error
Ang
le (d
eg)
(b) PDC
Figure 10 Angle tracking errors of knee and ankle joints
the coupling effect for a leg rehabilitation machine Experi-mental results show that the system response of the proposedapproach was in good agreement with that of the refer-ence input and guarantee robustness to system parameteruncertainties
Conflict of Interests
The author declares that there is no conflict of interestsregarding the publication of this paper
References
[1] D G Caldwell G A Medrano-Cerda and M GoodwinldquoControl of pneumatic muscle actuatorsrdquo IEEE Control SystemsMagazine vol 15 no 1 pp 40ndash48 1995
[2] C P Chou and B Hannaford ldquoStatic and dynamic characteris-tics ofMcKibben pneumatic artificialmusclesrdquo inProceedings ofthe IEEE International Conference on Robotics and Automationpp 281ndash286 May 1994
[3] T Noritsugu and T Tanaka ldquoApplication of rubber artificialmuscle manipulator as a rehabilitation robotrdquo IEEEASMETransactions on Mechatronics vol 2 no 4 pp 259ndash267 1997
12 The Scientific World Journal
[4] J H Lilly and L Yang ldquoSliding mode tracking for pneumaticmuscle actuators in opposing pair configurationrdquo IEEE Trans-actions on Control Systems Technology vol 4 pp 550ndash558 2005
[5] K K Ahn and H P H Anh ldquoDesign and implementation of anadaptive recurrent neural networks (ARNN) controller of thepneumatic artificial muscle (PAM) manipulatorrdquoMechatronicsvol 19 no 6 pp 816ndash828 2009
[6] X Shen ldquoNonlinearmodel-based control of pneumatic artificialmuscle servo systemsrdquo Control Engineering Practice vol 18 no3 pp 311ndash317 2010
[7] L A Zadeh ldquoFuzzy setsrdquo Information and Control vol 8 no 3pp 338ndash353 1965
[8] T Takagi and M Sugeno ldquoFuzzy identification of systems andits applications to modeling and controlrdquo IEEE Transactions onSystems Man and Cybernetics vol 15 no 1 pp 116ndash132 1985
[9] K K Ahn and H P H Anh ldquoInverse double NARX fuzzymodeling for system identificationrdquo IEEEASME Transactionson Mechatronics vol 15 no 1 pp 136ndash148 2010
[10] L Seddiki K Guelton and J Zaytoon ldquoConcept and Takagi-Sugeno descriptor tracking controller design of a closedmuscu-lar chain lower-limb rehabilitation devicerdquo IET Control Theoryand Applications vol 4 no 8 pp 1407ndash1420 2010
[11] K Tanaka T Ikeda and H O Wang ldquoFuzzy regulators andfuzzy observers relaxed stability conditions and LMI-baseddesignsrdquo IEEE Transactions on Fuzzy Systems vol 6 no 2 pp250ndash265 1998
[12] M K Chang J J Liou andM L Chen ldquoT-S fuzzymodel-basedtracking control of a one-dimensional manipulator actuated bypneumatic artificial musclesrdquo Control Engineering Practice vol19 no 12 pp 1442ndash1449 2011
[13] D B Reynolds D W Repperger C A Phillips and G BandryldquoModeling the dynamic characteristics of pneumatic musclerdquoAnnals of Biomedical Engineering vol 31 no 3 pp 310ndash3172003
[14] C S Tseng B S Chen and H J Uang ldquoFuzzy tracking controldesign for nonlinear dynamic systems via T-S fuzzy modelrdquoIEEE Transactions on Fuzzy Systems vol 9 no 3 pp 381ndash3922001
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
8 The Scientific World Journal
Table 2 Peak-peak error and phase lag for Figure 5
The proposed approach PDCPeak-peak error Phase lag Peak-peak error Phase lag
1205791
1205792
1205791
1205792
1205791
1205792
1205791
1205792
07 035 49∘ 45∘ 13 065 162∘ 108∘
20
10
0
minus10
minus20
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
ReferenceActual
15
0
minus15
1205792
(a) The proposed approach
20
10
0
minus10
minus20
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
ReferenceActual
15
0
minus15
1205792
(b) PDC
Figure 4 Sinusoidal wave response for knee and ankle joints
1198962= 001539 For comparison with the proposed controller
the PDC feedback gains are designed to be
1= [
minus03293 01023 minus02538 002317
minus01783 00021 minus002672 04732]
1198702= [
minus04278 01845 minus02451 00731
minus02185 00026 minus00752 00923]
1198703= [
minus06013 03461 minus05482 00231
minus04421 00093 minus00651 03529]
1198704= [
minus03756 03150 minus05391 00421
minus04250 00023 minus00531 03597]
(37)
51 Sinusoidal Wave Response Continuous reciprocation isrequired in order to foster the recovery of extremity functionThe sinusoidal wave responses of the proposed approach andPDC for both knee and ankle joints are shown in Figure 4It is evident that angle trajectories of the proposed approachare close to the command Figure 5 shows that the proposedapproach exhibits less tracking errors than does PDC Thepeak-peak error and phase lag are listed in Table 2 Because
of the interaction of the two joints PDC has significant angleerrors for 120579
1 which will degrade the rehabilitation effect
However supervisory control can overcome the couplingeffect of the two joints to achieve excellent rehabilitationfunction for patients
52 Irregular Curve Response In practical applications itcould be expected that the reference command will changewith different input frequencies The desired trajectories forboth knee and ankle joints are
1205791= 20 lowast 033 (sin (2120587119891
1119905) + sin (2120587119891
2119905) + sin (2120587119891
3119905))
1205792= 15 lowast 033 (sin (2120587119891
1119905) + sin (2120587119891
2119905) + sin (2120587119891
3119905))
(38)
with 1198911= 005Hz 119891
2= 01Hz and 119891
3= 0066Hz
Figure 6 shows the tracking responses of irregular curvesobtained using both the proposed approach and PDC Track-ing errors for the knee and ankle joints are shown in Figure 7Clearly the angle error of the proposed approach is averagemaintained within 2∘ However the proposed approach iscapable of adapting to different frequencies
The Scientific World Journal 9
14
12
10
8
6
4
2
0
minus2
minus41009080706050403020100
Time (s)
Ang
le (d
eg)
1205791 error1205792 error
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 5 Angle tracking errors for both knee and ankle
20
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
120579215
10
5
0
minus5
minus10
minus15
Actual
(a) The proposed approach
20
10
0
minus10
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
1205792
Actual
(b) PDC
Figure 6 Irregular curve response for both knee and ankle joints
53 Elliptic Response The desired end-effect or trajectory isgiven by
119909119889(119905) = 0614 minus 0015 sdot cos (02120587 sdot 119905 minus 120587)
119910119889(119905) = minus01 sdot sin (02120587 sdot 119905 minus 2120587)
(39)
where 0 le 119905 le 20 seconds
The end-effect tracking responses in the 119909 119910 coordinatefor both the proposed approach and PDC are shown inFigure 8 and the end-effect position tracking errors aredisplayed in Figure 9 It is evident that tracking behavior ofthe proposed approach is better than that of the PDC Ascan be seen the tracking errors of the proposed approach are
10 The Scientific World Journal
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 7 Angle tracking errors for both the proposed approach and PDC
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(a) The proposed approach
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(b) PDC
Figure 8 Elliptic response of the proposed approach and PDC
within 003m On the other hand angle tracking errors ofknee and ankle joints are shown in Figure 10
Moreover it is difficult to enhance end-effector trackingperformance using the PDC algorithm because the PDCcannot overcome the nonlinearity of PAMs and the structuralinteraction However the proposed approach overcomessuccessfully the coupling effect and parameter uncertaintiesof the system As seen in the experimental results theproposed approach can attain excellent end-effector trackingperformance in rehabilitation function
6 Conclusions
In this study a novel composite fuzzy control is proposed andapplied in the two-joint leg rehabilitation device driven byPAMsThe proposed controller is not only capable of decom-posing nonlinear systems into a set of linear subsystems but isalso capable of simplifying a complex nonlinear system usinglinear control techniques with the control gains determinedusingMATLABrsquos LMI Toolbox based on the Lyapunov stabil-ity theoremMoreover the supervisory control can overcome
The Scientific World Journal 11
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(a) The proposed approach
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(b) PDC
Figure 9 The end-effect position tracking errors
14
12
10
8
6
4
2
0
minus2
minus40 2 4 6 8 10 12 14 16 18 20
Ang
le (d
eg)
Time (s)
1205791 error1205792 error
(a) The proposed approach
0 2 4 6 8 10 12 14 16 18 20
Time (s)
15
10
5
0
minus5
minus10
1205791 error1205792 error
Ang
le (d
eg)
(b) PDC
Figure 10 Angle tracking errors of knee and ankle joints
the coupling effect for a leg rehabilitation machine Experi-mental results show that the system response of the proposedapproach was in good agreement with that of the refer-ence input and guarantee robustness to system parameteruncertainties
Conflict of Interests
The author declares that there is no conflict of interestsregarding the publication of this paper
References
[1] D G Caldwell G A Medrano-Cerda and M GoodwinldquoControl of pneumatic muscle actuatorsrdquo IEEE Control SystemsMagazine vol 15 no 1 pp 40ndash48 1995
[2] C P Chou and B Hannaford ldquoStatic and dynamic characteris-tics ofMcKibben pneumatic artificialmusclesrdquo inProceedings ofthe IEEE International Conference on Robotics and Automationpp 281ndash286 May 1994
[3] T Noritsugu and T Tanaka ldquoApplication of rubber artificialmuscle manipulator as a rehabilitation robotrdquo IEEEASMETransactions on Mechatronics vol 2 no 4 pp 259ndash267 1997
12 The Scientific World Journal
[4] J H Lilly and L Yang ldquoSliding mode tracking for pneumaticmuscle actuators in opposing pair configurationrdquo IEEE Trans-actions on Control Systems Technology vol 4 pp 550ndash558 2005
[5] K K Ahn and H P H Anh ldquoDesign and implementation of anadaptive recurrent neural networks (ARNN) controller of thepneumatic artificial muscle (PAM) manipulatorrdquoMechatronicsvol 19 no 6 pp 816ndash828 2009
[6] X Shen ldquoNonlinearmodel-based control of pneumatic artificialmuscle servo systemsrdquo Control Engineering Practice vol 18 no3 pp 311ndash317 2010
[7] L A Zadeh ldquoFuzzy setsrdquo Information and Control vol 8 no 3pp 338ndash353 1965
[8] T Takagi and M Sugeno ldquoFuzzy identification of systems andits applications to modeling and controlrdquo IEEE Transactions onSystems Man and Cybernetics vol 15 no 1 pp 116ndash132 1985
[9] K K Ahn and H P H Anh ldquoInverse double NARX fuzzymodeling for system identificationrdquo IEEEASME Transactionson Mechatronics vol 15 no 1 pp 136ndash148 2010
[10] L Seddiki K Guelton and J Zaytoon ldquoConcept and Takagi-Sugeno descriptor tracking controller design of a closedmuscu-lar chain lower-limb rehabilitation devicerdquo IET Control Theoryand Applications vol 4 no 8 pp 1407ndash1420 2010
[11] K Tanaka T Ikeda and H O Wang ldquoFuzzy regulators andfuzzy observers relaxed stability conditions and LMI-baseddesignsrdquo IEEE Transactions on Fuzzy Systems vol 6 no 2 pp250ndash265 1998
[12] M K Chang J J Liou andM L Chen ldquoT-S fuzzymodel-basedtracking control of a one-dimensional manipulator actuated bypneumatic artificial musclesrdquo Control Engineering Practice vol19 no 12 pp 1442ndash1449 2011
[13] D B Reynolds D W Repperger C A Phillips and G BandryldquoModeling the dynamic characteristics of pneumatic musclerdquoAnnals of Biomedical Engineering vol 31 no 3 pp 310ndash3172003
[14] C S Tseng B S Chen and H J Uang ldquoFuzzy tracking controldesign for nonlinear dynamic systems via T-S fuzzy modelrdquoIEEE Transactions on Fuzzy Systems vol 9 no 3 pp 381ndash3922001
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
The Scientific World Journal 9
14
12
10
8
6
4
2
0
minus2
minus41009080706050403020100
Time (s)
Ang
le (d
eg)
1205791 error1205792 error
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 5 Angle tracking errors for both knee and ankle
20
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
1205791
Ang
le (d
eg)
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
120579215
10
5
0
minus5
minus10
minus15
Actual
(a) The proposed approach
20
10
0
minus10
10
0
minus10
minus200 10 20 30 40 50 60 70 80 90 100
Time (s)
Ang
le (d
eg)
Ang
le (d
eg)
1205791
0 10 20 30 40 50 60 70 80 90 100
Time (s)
Reference
1205792
Actual
(b) PDC
Figure 6 Irregular curve response for both knee and ankle joints
53 Elliptic Response The desired end-effect or trajectory isgiven by
119909119889(119905) = 0614 minus 0015 sdot cos (02120587 sdot 119905 minus 120587)
119910119889(119905) = minus01 sdot sin (02120587 sdot 119905 minus 2120587)
(39)
where 0 le 119905 le 20 seconds
The end-effect tracking responses in the 119909 119910 coordinatefor both the proposed approach and PDC are shown inFigure 8 and the end-effect position tracking errors aredisplayed in Figure 9 It is evident that tracking behavior ofthe proposed approach is better than that of the PDC Ascan be seen the tracking errors of the proposed approach are
10 The Scientific World Journal
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 7 Angle tracking errors for both the proposed approach and PDC
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(a) The proposed approach
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(b) PDC
Figure 8 Elliptic response of the proposed approach and PDC
within 003m On the other hand angle tracking errors ofknee and ankle joints are shown in Figure 10
Moreover it is difficult to enhance end-effector trackingperformance using the PDC algorithm because the PDCcannot overcome the nonlinearity of PAMs and the structuralinteraction However the proposed approach overcomessuccessfully the coupling effect and parameter uncertaintiesof the system As seen in the experimental results theproposed approach can attain excellent end-effector trackingperformance in rehabilitation function
6 Conclusions
In this study a novel composite fuzzy control is proposed andapplied in the two-joint leg rehabilitation device driven byPAMsThe proposed controller is not only capable of decom-posing nonlinear systems into a set of linear subsystems but isalso capable of simplifying a complex nonlinear system usinglinear control techniques with the control gains determinedusingMATLABrsquos LMI Toolbox based on the Lyapunov stabil-ity theoremMoreover the supervisory control can overcome
The Scientific World Journal 11
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(a) The proposed approach
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(b) PDC
Figure 9 The end-effect position tracking errors
14
12
10
8
6
4
2
0
minus2
minus40 2 4 6 8 10 12 14 16 18 20
Ang
le (d
eg)
Time (s)
1205791 error1205792 error
(a) The proposed approach
0 2 4 6 8 10 12 14 16 18 20
Time (s)
15
10
5
0
minus5
minus10
1205791 error1205792 error
Ang
le (d
eg)
(b) PDC
Figure 10 Angle tracking errors of knee and ankle joints
the coupling effect for a leg rehabilitation machine Experi-mental results show that the system response of the proposedapproach was in good agreement with that of the refer-ence input and guarantee robustness to system parameteruncertainties
Conflict of Interests
The author declares that there is no conflict of interestsregarding the publication of this paper
References
[1] D G Caldwell G A Medrano-Cerda and M GoodwinldquoControl of pneumatic muscle actuatorsrdquo IEEE Control SystemsMagazine vol 15 no 1 pp 40ndash48 1995
[2] C P Chou and B Hannaford ldquoStatic and dynamic characteris-tics ofMcKibben pneumatic artificialmusclesrdquo inProceedings ofthe IEEE International Conference on Robotics and Automationpp 281ndash286 May 1994
[3] T Noritsugu and T Tanaka ldquoApplication of rubber artificialmuscle manipulator as a rehabilitation robotrdquo IEEEASMETransactions on Mechatronics vol 2 no 4 pp 259ndash267 1997
12 The Scientific World Journal
[4] J H Lilly and L Yang ldquoSliding mode tracking for pneumaticmuscle actuators in opposing pair configurationrdquo IEEE Trans-actions on Control Systems Technology vol 4 pp 550ndash558 2005
[5] K K Ahn and H P H Anh ldquoDesign and implementation of anadaptive recurrent neural networks (ARNN) controller of thepneumatic artificial muscle (PAM) manipulatorrdquoMechatronicsvol 19 no 6 pp 816ndash828 2009
[6] X Shen ldquoNonlinearmodel-based control of pneumatic artificialmuscle servo systemsrdquo Control Engineering Practice vol 18 no3 pp 311ndash317 2010
[7] L A Zadeh ldquoFuzzy setsrdquo Information and Control vol 8 no 3pp 338ndash353 1965
[8] T Takagi and M Sugeno ldquoFuzzy identification of systems andits applications to modeling and controlrdquo IEEE Transactions onSystems Man and Cybernetics vol 15 no 1 pp 116ndash132 1985
[9] K K Ahn and H P H Anh ldquoInverse double NARX fuzzymodeling for system identificationrdquo IEEEASME Transactionson Mechatronics vol 15 no 1 pp 136ndash148 2010
[10] L Seddiki K Guelton and J Zaytoon ldquoConcept and Takagi-Sugeno descriptor tracking controller design of a closedmuscu-lar chain lower-limb rehabilitation devicerdquo IET Control Theoryand Applications vol 4 no 8 pp 1407ndash1420 2010
[11] K Tanaka T Ikeda and H O Wang ldquoFuzzy regulators andfuzzy observers relaxed stability conditions and LMI-baseddesignsrdquo IEEE Transactions on Fuzzy Systems vol 6 no 2 pp250ndash265 1998
[12] M K Chang J J Liou andM L Chen ldquoT-S fuzzymodel-basedtracking control of a one-dimensional manipulator actuated bypneumatic artificial musclesrdquo Control Engineering Practice vol19 no 12 pp 1442ndash1449 2011
[13] D B Reynolds D W Repperger C A Phillips and G BandryldquoModeling the dynamic characteristics of pneumatic musclerdquoAnnals of Biomedical Engineering vol 31 no 3 pp 310ndash3172003
[14] C S Tseng B S Chen and H J Uang ldquoFuzzy tracking controldesign for nonlinear dynamic systems via T-S fuzzy modelrdquoIEEE Transactions on Fuzzy Systems vol 9 no 3 pp 381ndash3922001
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
10 The Scientific World Journal
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
(a) The proposed approach
Ang
le (d
eg)
1009080706050403020100
Time (s)
1205791 error1205792 error
15
10
5
0
minus5
minus10
(b) PDC
Figure 7 Angle tracking errors for both the proposed approach and PDC
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(a) The proposed approach
02
015
01
005
0
minus005
minus01
minus015
minus020 01 02 03 04 05 06 07 08
Y(m
)
X (m)
ReferenceActual
Initialposition
(b) PDC
Figure 8 Elliptic response of the proposed approach and PDC
within 003m On the other hand angle tracking errors ofknee and ankle joints are shown in Figure 10
Moreover it is difficult to enhance end-effector trackingperformance using the PDC algorithm because the PDCcannot overcome the nonlinearity of PAMs and the structuralinteraction However the proposed approach overcomessuccessfully the coupling effect and parameter uncertaintiesof the system As seen in the experimental results theproposed approach can attain excellent end-effector trackingperformance in rehabilitation function
6 Conclusions
In this study a novel composite fuzzy control is proposed andapplied in the two-joint leg rehabilitation device driven byPAMsThe proposed controller is not only capable of decom-posing nonlinear systems into a set of linear subsystems but isalso capable of simplifying a complex nonlinear system usinglinear control techniques with the control gains determinedusingMATLABrsquos LMI Toolbox based on the Lyapunov stabil-ity theoremMoreover the supervisory control can overcome
The Scientific World Journal 11
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(a) The proposed approach
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(b) PDC
Figure 9 The end-effect position tracking errors
14
12
10
8
6
4
2
0
minus2
minus40 2 4 6 8 10 12 14 16 18 20
Ang
le (d
eg)
Time (s)
1205791 error1205792 error
(a) The proposed approach
0 2 4 6 8 10 12 14 16 18 20
Time (s)
15
10
5
0
minus5
minus10
1205791 error1205792 error
Ang
le (d
eg)
(b) PDC
Figure 10 Angle tracking errors of knee and ankle joints
the coupling effect for a leg rehabilitation machine Experi-mental results show that the system response of the proposedapproach was in good agreement with that of the refer-ence input and guarantee robustness to system parameteruncertainties
Conflict of Interests
The author declares that there is no conflict of interestsregarding the publication of this paper
References
[1] D G Caldwell G A Medrano-Cerda and M GoodwinldquoControl of pneumatic muscle actuatorsrdquo IEEE Control SystemsMagazine vol 15 no 1 pp 40ndash48 1995
[2] C P Chou and B Hannaford ldquoStatic and dynamic characteris-tics ofMcKibben pneumatic artificialmusclesrdquo inProceedings ofthe IEEE International Conference on Robotics and Automationpp 281ndash286 May 1994
[3] T Noritsugu and T Tanaka ldquoApplication of rubber artificialmuscle manipulator as a rehabilitation robotrdquo IEEEASMETransactions on Mechatronics vol 2 no 4 pp 259ndash267 1997
12 The Scientific World Journal
[4] J H Lilly and L Yang ldquoSliding mode tracking for pneumaticmuscle actuators in opposing pair configurationrdquo IEEE Trans-actions on Control Systems Technology vol 4 pp 550ndash558 2005
[5] K K Ahn and H P H Anh ldquoDesign and implementation of anadaptive recurrent neural networks (ARNN) controller of thepneumatic artificial muscle (PAM) manipulatorrdquoMechatronicsvol 19 no 6 pp 816ndash828 2009
[6] X Shen ldquoNonlinearmodel-based control of pneumatic artificialmuscle servo systemsrdquo Control Engineering Practice vol 18 no3 pp 311ndash317 2010
[7] L A Zadeh ldquoFuzzy setsrdquo Information and Control vol 8 no 3pp 338ndash353 1965
[8] T Takagi and M Sugeno ldquoFuzzy identification of systems andits applications to modeling and controlrdquo IEEE Transactions onSystems Man and Cybernetics vol 15 no 1 pp 116ndash132 1985
[9] K K Ahn and H P H Anh ldquoInverse double NARX fuzzymodeling for system identificationrdquo IEEEASME Transactionson Mechatronics vol 15 no 1 pp 136ndash148 2010
[10] L Seddiki K Guelton and J Zaytoon ldquoConcept and Takagi-Sugeno descriptor tracking controller design of a closedmuscu-lar chain lower-limb rehabilitation devicerdquo IET Control Theoryand Applications vol 4 no 8 pp 1407ndash1420 2010
[11] K Tanaka T Ikeda and H O Wang ldquoFuzzy regulators andfuzzy observers relaxed stability conditions and LMI-baseddesignsrdquo IEEE Transactions on Fuzzy Systems vol 6 no 2 pp250ndash265 1998
[12] M K Chang J J Liou andM L Chen ldquoT-S fuzzymodel-basedtracking control of a one-dimensional manipulator actuated bypneumatic artificial musclesrdquo Control Engineering Practice vol19 no 12 pp 1442ndash1449 2011
[13] D B Reynolds D W Repperger C A Phillips and G BandryldquoModeling the dynamic characteristics of pneumatic musclerdquoAnnals of Biomedical Engineering vol 31 no 3 pp 310ndash3172003
[14] C S Tseng B S Chen and H J Uang ldquoFuzzy tracking controldesign for nonlinear dynamic systems via T-S fuzzy modelrdquoIEEE Transactions on Fuzzy Systems vol 9 no 3 pp 381ndash3922001
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
The Scientific World Journal 11
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(a) The proposed approach
02
018
016
014
012
01
008
006
004
002
00 2 4 6 8 10 12 14 16 18 20
Time (s)
Posit
ion
(m)
(b) PDC
Figure 9 The end-effect position tracking errors
14
12
10
8
6
4
2
0
minus2
minus40 2 4 6 8 10 12 14 16 18 20
Ang
le (d
eg)
Time (s)
1205791 error1205792 error
(a) The proposed approach
0 2 4 6 8 10 12 14 16 18 20
Time (s)
15
10
5
0
minus5
minus10
1205791 error1205792 error
Ang
le (d
eg)
(b) PDC
Figure 10 Angle tracking errors of knee and ankle joints
the coupling effect for a leg rehabilitation machine Experi-mental results show that the system response of the proposedapproach was in good agreement with that of the refer-ence input and guarantee robustness to system parameteruncertainties
Conflict of Interests
The author declares that there is no conflict of interestsregarding the publication of this paper
References
[1] D G Caldwell G A Medrano-Cerda and M GoodwinldquoControl of pneumatic muscle actuatorsrdquo IEEE Control SystemsMagazine vol 15 no 1 pp 40ndash48 1995
[2] C P Chou and B Hannaford ldquoStatic and dynamic characteris-tics ofMcKibben pneumatic artificialmusclesrdquo inProceedings ofthe IEEE International Conference on Robotics and Automationpp 281ndash286 May 1994
[3] T Noritsugu and T Tanaka ldquoApplication of rubber artificialmuscle manipulator as a rehabilitation robotrdquo IEEEASMETransactions on Mechatronics vol 2 no 4 pp 259ndash267 1997
12 The Scientific World Journal
[4] J H Lilly and L Yang ldquoSliding mode tracking for pneumaticmuscle actuators in opposing pair configurationrdquo IEEE Trans-actions on Control Systems Technology vol 4 pp 550ndash558 2005
[5] K K Ahn and H P H Anh ldquoDesign and implementation of anadaptive recurrent neural networks (ARNN) controller of thepneumatic artificial muscle (PAM) manipulatorrdquoMechatronicsvol 19 no 6 pp 816ndash828 2009
[6] X Shen ldquoNonlinearmodel-based control of pneumatic artificialmuscle servo systemsrdquo Control Engineering Practice vol 18 no3 pp 311ndash317 2010
[7] L A Zadeh ldquoFuzzy setsrdquo Information and Control vol 8 no 3pp 338ndash353 1965
[8] T Takagi and M Sugeno ldquoFuzzy identification of systems andits applications to modeling and controlrdquo IEEE Transactions onSystems Man and Cybernetics vol 15 no 1 pp 116ndash132 1985
[9] K K Ahn and H P H Anh ldquoInverse double NARX fuzzymodeling for system identificationrdquo IEEEASME Transactionson Mechatronics vol 15 no 1 pp 136ndash148 2010
[10] L Seddiki K Guelton and J Zaytoon ldquoConcept and Takagi-Sugeno descriptor tracking controller design of a closedmuscu-lar chain lower-limb rehabilitation devicerdquo IET Control Theoryand Applications vol 4 no 8 pp 1407ndash1420 2010
[11] K Tanaka T Ikeda and H O Wang ldquoFuzzy regulators andfuzzy observers relaxed stability conditions and LMI-baseddesignsrdquo IEEE Transactions on Fuzzy Systems vol 6 no 2 pp250ndash265 1998
[12] M K Chang J J Liou andM L Chen ldquoT-S fuzzymodel-basedtracking control of a one-dimensional manipulator actuated bypneumatic artificial musclesrdquo Control Engineering Practice vol19 no 12 pp 1442ndash1449 2011
[13] D B Reynolds D W Repperger C A Phillips and G BandryldquoModeling the dynamic characteristics of pneumatic musclerdquoAnnals of Biomedical Engineering vol 31 no 3 pp 310ndash3172003
[14] C S Tseng B S Chen and H J Uang ldquoFuzzy tracking controldesign for nonlinear dynamic systems via T-S fuzzy modelrdquoIEEE Transactions on Fuzzy Systems vol 9 no 3 pp 381ndash3922001
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
12 The Scientific World Journal
[4] J H Lilly and L Yang ldquoSliding mode tracking for pneumaticmuscle actuators in opposing pair configurationrdquo IEEE Trans-actions on Control Systems Technology vol 4 pp 550ndash558 2005
[5] K K Ahn and H P H Anh ldquoDesign and implementation of anadaptive recurrent neural networks (ARNN) controller of thepneumatic artificial muscle (PAM) manipulatorrdquoMechatronicsvol 19 no 6 pp 816ndash828 2009
[6] X Shen ldquoNonlinearmodel-based control of pneumatic artificialmuscle servo systemsrdquo Control Engineering Practice vol 18 no3 pp 311ndash317 2010
[7] L A Zadeh ldquoFuzzy setsrdquo Information and Control vol 8 no 3pp 338ndash353 1965
[8] T Takagi and M Sugeno ldquoFuzzy identification of systems andits applications to modeling and controlrdquo IEEE Transactions onSystems Man and Cybernetics vol 15 no 1 pp 116ndash132 1985
[9] K K Ahn and H P H Anh ldquoInverse double NARX fuzzymodeling for system identificationrdquo IEEEASME Transactionson Mechatronics vol 15 no 1 pp 136ndash148 2010
[10] L Seddiki K Guelton and J Zaytoon ldquoConcept and Takagi-Sugeno descriptor tracking controller design of a closedmuscu-lar chain lower-limb rehabilitation devicerdquo IET Control Theoryand Applications vol 4 no 8 pp 1407ndash1420 2010
[11] K Tanaka T Ikeda and H O Wang ldquoFuzzy regulators andfuzzy observers relaxed stability conditions and LMI-baseddesignsrdquo IEEE Transactions on Fuzzy Systems vol 6 no 2 pp250ndash265 1998
[12] M K Chang J J Liou andM L Chen ldquoT-S fuzzymodel-basedtracking control of a one-dimensional manipulator actuated bypneumatic artificial musclesrdquo Control Engineering Practice vol19 no 12 pp 1442ndash1449 2011
[13] D B Reynolds D W Repperger C A Phillips and G BandryldquoModeling the dynamic characteristics of pneumatic musclerdquoAnnals of Biomedical Engineering vol 31 no 3 pp 310ndash3172003
[14] C S Tseng B S Chen and H J Uang ldquoFuzzy tracking controldesign for nonlinear dynamic systems via T-S fuzzy modelrdquoIEEE Transactions on Fuzzy Systems vol 9 no 3 pp 381ndash3922001
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of