Professional Documents
Culture Documents
Mojtaba Hadi Barhaghtalab, Hassan Bayani*, Armin Nabaei, Houman Zarrabi, Ali Amiri
In this study, a robust neuro-adaptive controller for cable-driven parallel robots is proposed. The robust neuro-
adaptive control system is comprised of a computation controller and a robust controller. The computation controller
containing a neural-network-estimator with radial basis function activator is the principal controller and the robust
controller is designed to achieve tracking performance. An on-line tuning method is derived to tune the parameters
of the neural network for estimating the controlled system dynamic function. To investigate the effectiveness of
the robust adaptive control, the design methodology is applied to control a cable-driven parallel robot. Simulation
results demonstrate that the proposed robust adaptive control system can achieve favorable tracking performances
for the robot.
Key words: Robust control, Adaptive law, neural network, Cable-driven Parallel Robot
Dizajn robusnog neuro-adaptivnog regulatora za žično pogonjene paralelne robote. U ovome redu pred-
stavljen je neuro-adaptivni regulator za žično pogonjene paralelne robote. Robusni neuro-adaptivni regulator sastoji
se od regulatora zasnovanog na estimiranom modelu i robusnog regulatora. Prvi regulator sadrži estimator s neuron-
skom mrežom s radijalnom aktivacijskom funkcijom glavni je regulator u sustavu, a robusni je regulator dizajniran
za slijed̄enje. Izvedena je on-line metoda podešavanja parametara neuronske mreže za estimaciju dinamike sustava
upravljanja. Efikasnost sustava adaptivnog, robusnog regulatora testirana je na na žično pogonjenom paralelnom
robotu. Simulacijski rezultati pokazuju da se predloženim robusnim i adaptivnim regulatorom mogu dobiti zadovo-
ljavajuće performanse prilikom slijed̄enja.
Ključne riječi: robusno upravljanje, adaptivni upravljački zakon, neuronska mreža, žično pogonjeni paralelni ro-
bot
1 INTRODUCTION of CDPR for real-time purposes [6, 7]. For all kinetostatic
In the recent decades, robots have been utilized in vast model proposed for CDPR, there should be a static balan-
area of industries. However, many parts of manufactu- cing between external forces and tension of the cables [8].
ring and engineering does not put upon robots, mainly due With assumption of no mass and no elongation for cables,
to the weakness of conventional robots [1] [2]. For ins- most of the common control strategies used for conventi-
tance, in many applications workspace requirements and onal robots could be adapted for CDPRs. In case of cla-
load carrying capacity are so much higher than what the ssic controller, PD controller is applied in [9] and the re-
conventional robots can provide while cost of the robot sults present acceptable performance over desired task. As
should be considered [3] [4]. Toward resolving the lat- modern control approaches, inverse control of cable-driven
ter issue, new class of parallel robots were introduced [5]. parallel mechanism using ANFIS is the main contribution
Cable-Driven Parallel Robots (CDPR) are structurally si- in [10]. Nonlinear Sliding Mode Controller (SMC) and fe-
milar to parallel actuated robots but with the fundamental asible workspace analysis for a cable suspended robot with
difference that cables can only pull the End-Effector (EE) input constraints is presented in [11].
but not push it. Figure 1 represents schematically a CDPR All the above efforts for deriving kinematic and dynamic
in a general arrangement. It consists of motor, winch sys- equations tend to present a simple model of CDPR that
tem and the EE. From a scientific stand point, feedback could works in an online manner with common control ap-
control of CDPR is lot more challenging than their coun- proaches while precision is not devastated [12, 13]. Many
terpart parallel-actuated robots due to the cables behavior. researches illustrates effectiveness of the precise modeling
Several efforts had been exerted on modeling and control of the robot in the controlling procedure [14, 15], howe-
A2 Am
A1
l2
lm
l1
S2 z
ai B2 Sm
S1 y Bm
b1
B1 x
OE
Z c
Slika 1. Schematic representation of a CDPR with m ca-
bles.
Y
X OG
ver controller capabilities may affect the tracking perfor-
mance [16, 17]. The main contribution of this paper, com-
pared to similar researches toward controlling of CDPR Slika 2. kinematic modeling of a CDPR with m cables.
is taking benefits of Robust Neuro-Adaptive Controller
(RNAC). The latter method allows the design of considera-
bly more precise, energy-efficient and compliant controls vectors in OG and OE coordinates respectively. Note that
for robots. It allows one to obtain a linear closed-loop equ- OG is the global coordinate and OE is the attached coordi-
ation in terms of the state variables [18, 19]. This fact has nates on the EE. Referring to Fig. 2, the following equation
no precedent in the study of the controllers. On the ot- is derived [20]:
her hand, RNAC is characterized for being a dynamic con- c = ai + li − bi (1)
troller, that is, its complete control law includes additional
state variables. Moreover, as another contribution radial Equation (1) could be written in the OG coordinates as fol-
basis function as an activator function is used in the struc- lows:
ture of the neural network part of the adaptive law. c = ai + li − R × bi (2)
The organization of this paper is as follows. Section II which could be rearranged in the following form:
describes the kinematic, static and dynamic equations of
CDPR. The general forms of these equations are derived li = c − ai + R × bi (3)
for any structure of the CDPR. Section III and IV outline
a method for control based on RNAC which contains of where R is the transformation matrix from OE to OG
obtaining the control rule. Finally, the paper concludes by which could be written as follows:
presenting the simulation results and some hints for ongo-
cψ cφ cψ s1 sψ − cθ sψ cθ cψ sφ + sθ sψ
ing works. R = cψ sψ cθ cψ + sθ sφ sψ −cψ sθ + cθ sφ sψ (4)
−sφ cφ sθ cφ cθ
2 CABLE-DRIVEN PARALLEL ROBOTS MODEL
In Eq. (4), c stands for cos(·) and s indicates sin(·). Also
This section outlines the kinematic, static and dynamic θ , φ and ψ are angle of rotation around x, y and z axis
of the CDPR in general arrangement. The presented model respectively. By taking the first derivative of Eq. (3) the
is based on [20]. following equation is obtained:
l̇i = Si T ċ + (bi × Si T )ω
ωi (7)
Hc = Icω (17) Slika 4. Parameters of the under study CDPR with eight
cables.
where Ic denotes the EE moment of inertia around its mass
center. By taking the first derivative of Eq. (17) the fol-
lowing equation is derived: where G = [0 0 9.81 0 0 0]T is the gravity vector, q is the
∂ Hc vector of position, M(q) denoted the mass matrix and is
Ḣc = ω × Hc = I˙cω + Icω̇
+ ω̇ ω +ω ω Icω = Icω̇
ω +ω ω Icω defined as follows [20] [22]:
∂t
(18) mI3×3 03×3
By using Eq. (17) and Euler equation, one could derive the M(q) = (24)
03×3 PT IP P
following equation:
where
n
ω Icω = ∑ bi × Siτ i
ω +ω
Icω̇ (19) IP = RIRT (25)
i=1 In addition, C(q,q̇) the so-called coriolis terms and centri-
fugal forces and J(q) denoted the Jacobian matrix of the
Note that ω for the point c is defined as follows:
CDPR illustrated in Fig. 4, which compromise eight ca-
bles, are defined as follows:
ω1 θ̇
ω2 = P φ̇
(20) 0
ω3 ψ̇ 0
C(q,q̇) =
(26)
0
where ω c = [θ̇ φ̇ ψ̇] is the angular velocity of the mass pIṖq̇ + Ṗq̇ × IṖq̇
center in EE and P is defined as follows:
I3×3 03×3
cφ cψ −sψ 0 J(q) = (27)
03×3 P
P = cφ sψ cψ 0 (21)
−sφ 0 1 The following equation represents the dynamic of the pul-
ley [22]:
Therefor the angular acceleration could be resolved from Nz θ̈ + Cθ̇ = τ (28)
the following equation: where Nz is the matrix of pulley inertia with respect to
z − axis, θ is angel of each pulley and C is rotational vis-
α1 cous damping matrix which are defined due to the charac-
α2 = Ṗω̇ω c + Pω̈
ωc (22) teristics of the motor and system of winch and could be
α3 expressed as follows:
where αi , i = {1, 2, 3} denotes the ith components of an- J1 0 0 0 C1 0 0 0
gular acceleration. Using Eqs. (16) to (22), the following 0 J2 0 0 0 C2 0 0
J=
0 0 J3 0 , C = 0
dynamic equation for the CDPR is derived: 0 C3 0
0 0 0 J4 0 0 0 C4
− JT(q)τ = M(q) q̈ + C(q,q̇) q̇ + M(q) G (23) (29)
Based on the kinematic, static and dynamic equations de- where ∆ denotes an approximation error, w∗ and Φ∗ are
rived here, the following part investigates the proposed the optimal parameter vectors of w and Φ, respectively,
controller for the eight cables CDPR. Next part provides and m∗ and s∗ are the optimal parameter vectors of m and
the neural network estimator for the dynamic model with s, respectively. The optimal weighting vectors w∗ , m∗ and
Taylor linearization method. s∗ which are needed to best approximate a given nonlinear
function Θ are difficult to determine. An estimated RBF
3 RADIAL BASIS FUNCTION NEURAL approximator is defined as the following form [29]:
NETWORK
ŷ = ŵT Φ̂(x, m̂, ŝ) (37)
One of the promising estimators that could operates in
real-time simulations are neural network with radial basis where ŵ and Φ̂ are the estimated vectors of w∗ and Φ∗ ,
functions as the activation functions. They are highly accu- respectively, and m̂ and ŝ are the estimated vectors of m∗
rate estimator and they could simulate any model including and s∗ , respectively. The estimated error ỹ is defined as:
nonlinearities. A simple RBF neural network maps can be
written according to the following equation [25]: ỹ = Θ − ŷ = y∗ − ŷ + ∆
theorem, there exists an optimal RBF neural network such control system is designed as Eq. (51), in which the adapta-
that [28]: tion laws of the system dynamic function approximator are
f (φ , φ ) = f ∗ (`∗ ,t) + ε (47) designed as in Eq. (56) and Eq. (58), and the robust con-
troller is designed as in Eq. (59). Henceforth, the stability
where θ ∗ = [w∗ m∗ s∗ ]T is the optimal weight vector of
of the system can be guaranteed [32]. The latter statement
the RBF neural network and ε denotes the approximation
could be regarded as follows:
error. However, the optimal weight vector is difficult to
determine. Define the weight estimated error vector ˜` as: ˙ = −η1 eT PBm Φ̂
ŵ = −w̃ (56)
θ̃ = θ ∗ − θ̂ (48) m̂ = −m̃ ˙ = −η2 eT PBm Aŵ (57)
where θ̂ = [ŵ m̂ ŝ ]T is the estimated vector of the optimal ŝ = −s̃˙ = −η3 eT PBm Bŵ (58)
vector θ ∗ . While ε appears, the following H ∞ tracking 1 T
uR = B Pe (59)
performance is requested [32]: κ m
Z T Z T where η1 , η2 and η3 are the learning rates with positive
1 T constants, κ is a positive weighting factor, and positive
eT Qedt ≤ eT (0)Pe(0) + θ̃ (0)θ̃(0) + ρ 2 ε2 dt
0 κ 0 matrix P = PT is the solution of the following Riccati-like
(49) equation:
where T ∈ [0, ∞] and ε ∈ L2 [0, T ]. The Q = QT and P = PT
are given positive weighting matrices, κ is a design gain, 2 1
PAm + ATm P + Q − PBm BTm P + 2 PBm BTm P = 0 (60)
and ρ is a prescribed attenuation level. If the system starts κ ρ
with initial conditions e(0) = 0 and θ̃(0) = 0, then the H ∞
if and only if κ2 − ρ12 ≥ 0 or 2ρ 2 ≥ κ [35-37]. Therefore, for
performance in Eq. (49) can be rewritten as [33, 34]:
a prescribed ρ in H ∞ tracking control, in order to guaran-
RT T
0 e Qe dt
tee the solvability of H ∞ tracking performance, the weight
sup RT ≤ρ (50) κ on control law uR of Eq. (59) should satisfy the above
ε∈L2 [0,T ] ”2 dt
0
inequality. Then, the H ∞ tracking performance in Eq. (49)
where the L2 -gain from ε to the tracking error e should can be achieved for a prescribed attenuation level ρ. Afo-
be equal or less than ρ. In order to achieving a favorable rementioned, the robust neuro-adaptive controller includes
tracking performance and an arbitrarily small attenuation three different controlling efforts. These include the efforts
level simultaneously, a block diagram for robust adaptive to create stability by Lyapunov criterion, zero error and ap-
controller is shown in Fig. 5. proximation of the system uncertainties. Robust controller
approximate the uncertainties of the model while they are
uRF = uCP + uR (51) also decreased using RBF estimator. The robust adaptive
controller design procedure is shown in algorithm 2 in ge-
in which the computation controller is chosen as [35]: neral. In the first line, constants of neural network and
uCP = − fˆNN + φ̈m + k2 ė + k1 e (52) robust controller are determined. Next, in the second line,
the desired signal and its derivative are obtained. The third
with the dynamic approximator chosen as [36]: line consist of computing the new models based on the in-
put and output of the model. Based on defined error control
fˆNN (e, m̂, ŝ, ŵ) = ŵT Φ̂(e, m̂, ŝ). (53) process begins. In lines 5 to 8, if five steps of identifying
new model is past, the new model will be used to imple-
The estimated vectors ŵ, m̂ and ŝ are the optimal vectors of
ment the controller; otherwise, the previous model is used.
w∗ , m∗ and s∗ , respectively. By substituting Eq. (51) into
In the ninth line, the error vector of the identification is
Eq. (44) and using Eq. (45), the tracking error dynamic
determined and based on this vector; future estimation for
equation can be obtained as follows [33]:
more accurate model is obtained. In the line ten, error of
ė = Am e − Bm ( f − fˆNN + uR ) (54) the controller signal is determined based Herwitz equation.
The thirteenth line determines the controller output errors.
0 1 T The P matrix is determined for the next step. This pro-
where Am = , and Bm = 0 1 . Using
−k1 −k2 cedure continues until achievement of the desired control.
Eq. (43), Eq. (54) can be rewritten as: The parameters bi for the dynamic coefficients in Eq. (44)
are given in [32]. Solving the Riccati-like equation repre-
ė = Am e − Bm (w̃T Φ̂ + m̃T Aŵ + s̃T Bŵ + ε + uR ) (55) sented in Eq. (60) with ρ 2 = κ, one has:
Consider the proposed motion equations which are repre- −1 0
sented by Eq. (23) and (43). If the robust adaptive fuzzy Q= (61)
0 −1
and
1.8 0.8
P= (62)
0.8 0.8
It should be emphasized that the derivation of controller
equations does not need to use the dynamic parameters and
the structure of the dynamic functions. The system para-
meters are used only for simulations. In order to inves-
tigate the effectiveness of the developed control system,
initial conditions are simulated. The RBF neural network
with seven neurons in hidden layer is utilized to approach
the under study two-CDPR dynamics. The learning rates
are selected as:
In order to attenuate to a small level via H ∞ tracking design Slika 7. Desired and operated path for displacement in
technique, the simulation results of the controller with κ = y-direction.
0.1 are derived.
Slika 8. Step response for displacement in z-direction. Slika 10. Step response for rotation around y-direction (φ ).
Slika 9. Step response for rotation around x-direction (θ ). Slika 11. Step response for rotation around z-direction (ψ).
robot. In addition, the stability of the proposed controller [3] S. Kawamura, W. Choe, S. Tanaka, and S. R Pandian. De-
was verified. Simulation results have shown effectiveness velopment of an Ultrahigh Speed Robot Falcon Using Wire
of this controller. Finally, the developed robust Neuro- Drive System., International Conference on Robotics and
adaptive control system was applied to control a CDPR, Automation, Proceedings of IEEE,1995, volume 1, pp.215–
which demonstrates suitable performance. Ongoing works 220. IEEE, 1995.
includes implementation of the proposed controller for real
robot using vision procedure. [4] Hassan Bayani, et, al. “On the control of planar cable-driven
parallel robot via classic controllers and tuning with intelli-
Literatura gent algorithms ”, Third RSI/ISM International Conference
on Robotics and Mechatronics (ICRoM), pp. 623-628, 2015.
[1] P. Dewdney, et, al. “The Large Adaptive Reflector: A Gi-
ant Radio Telescope With An Aero Twist,”, Canadian Aero- [5] Guo, Hongbo, et al. “Cascade control of a hydraulically
nautics journal,2002. driven 6-DOF parallel robot manipulator based on a sliding
mode,” Control Engineering Practice 16.9 (2008): pp.1055-
[2] Hassan Bayani, et, al. “On the determination of the maxi-
1068.
mal inscribed ellipsoid in the Wrench-Feasible Workspace of
the cable-driven parallel robots”, Second RSI/ISM Interna- [6] J. Brest,et al. Self-Adapting Control Parameters in Differen-
tional Conference on Robotics and Mechatronics (ICRoM), tial Evolution: A Comparative Study on Numerical Bench-
pp. 422-427, 2014. mark Problems, IEEE Transaction on computation, 2006.
Slika 12. The frequency response of the designed controller (a) Cables one to four tension.
in constrained motion.
Cable Robot." In Cable-Driven Parallel Robots, pp. 337-352. [32] Fradkov, A.L., Miroshnik, I.V. and Nikiforov, V.O., " No-
Springer, 2013. nlinear and adaptive control of complex systems", Springer
[19] Z. Jinhua , Z. Jian, D. Haifeng, W. Sunan, Self-organizing Science and Business Media, Vol. 491, 2013.
genetic algorithm based tuning of PID controllers, Journal
[33] Vukobratovic, Miomir, Dragan Stokic, and Nenad Kircan-
of Information Sciences, 2009, VOL.179, NO.2, pp.1007–
ski. "Non-adaptive and adaptive control of manipulation
1018.
robots",Springer Science and Business Media, Vol. 5, 2013.
[20] Reza Babaghasabha, Mohammad A. Khosravi and Hamid
D. Taghirad. "Adaptive Control of KNTU Planar Cable- [34] Fateh, Mohammad Mehdi, and Saeed Khorashadizadeh.
Driven Parallel Robot with Uncertainties in Dynamic and "Robust control of electrically driven robots by adaptive fu-
Kinematic Parameters",Mechanisms and Machine Science, zzy estimation of uncertainty", Nonlinear Dynamics, Vol.69,
2011. No.3, pp. 1465-1477, 2012.
[21] Guo, Hongbo, et al. “Cascade control of a hydraulically
driven 6-DOF parallel robot manipulator based on a sliding [35] Schenk, Christian, Heinrich H. Bulthoff, and Carlo Masone.
mode,” Control Engineering Practice 16.9 (2008): pp.1055- "Robust adaptive sliding mode control of a redundant cable
1068. driven parallel robot", 19th International Conference on Sys-
tem Theory, Control and Computing (ICSTCC), 2015.
[22] Korayem, Moharam Habibnejad and Tourajizadeh, H
“Maximum DLCC of Spatial Cable Robot For a Predefined [36] Salas-Pena, Oscar, Herman Castaneda, and Jesus de Leon-
Trajectory Within The Workspace Using Closed Loop Opti- Morales. "Robust Adaptive Control for a DC Servomotor
mal Control Approach’, Journal of Intelligent & Robotic Sys- with wide Backlash Nonlinearity", Automatika: Journal for
tems,springer, vol. 63, no. 1, pp. 75–99, 2011. Control, Measurement, Electronics, Computing and Com-
[23] Middleton, RH and Goodwin, GC "Adaptive computed munications, Vol.56, No.4, 2015.
torque control for rigid link manipulations", Elsevier,1988.
[24] Yang, Zhiyong, Jiang Wu, and Jiangping Mei. "Motor- [37] Kuljaca, Ognjen, Krunoslav Horvat, and Bruno Borovic.
mechanism dynamic model based neural network optimized "Design of adaptive neural network controller for thermal
computed torque control of a high speed parallel manipula- power system frequency control", Automatika: Journal for
tor."Mechatronics 17.7 (2007). Control, Measurement, Electronics, Computing and Com-
munications, Vol.52, No.4, 2012.
[25] Reiner, Philip David. "Algorithms for Optimal Cons-
truction and Training of Radial Basis Function Neural
Networks", Diss. Auburn University, 2015.
M.H.Barhaghtalab Mojtaba Hadi Barhaghtalab
[26] Raj, Devi R., I. Jacob Raglend, and M. Anand. "Inverse received his B.Sc. degree in Electrical Engine-
kinematics solution of a five joint robot using feed forward ering (Communication orientation) from Islamic
and Radial Basis Function Neural Network", International Azad University, Science and Research Branch,
Conference on Computation of Power, Energy Information Iran (2012) and M.Sc. degree in Electrical Engi-
and Commuincation (ICCPEIC), 2015. neering (Control orientation) from Persian Gulf
University, Bushehr, Iran (2016). His main rese-
[27] Daachi, M.E., Madani, T., Daachi, B. and Djouani, K., "A arch interests include robotic, intelligent system,
radial basis function neural network adaptive controller to fuzzy control, neural network, adaptive control,
drive a powered lower limb knee joint orthosis.", Applied nonlinear control, stability analysis of systems.
Soft Computing, pp.324-336. , 2015.
[28] Reiner, Philip David. "Algorithms for Optimal Cons-
truction and Training of Radial Basis Function Neural H.Bayani received the B.S. Eng. Degree in Mec-
Networks", Diss. Auburn University, 2015. hanical Engineering from the Iran University of
[29] Huang, Hsu-Chih, and Chih-Hao Chiang. "An Evolutionary Science and Technology (IUST) Tehran, Iran, in
Radial Basis Function Neural Network with Robust Genetic- 2011. He worked on designing & building of a
model helicopter as Bachelor degree thesis de-
Based Immunecomputing for Online Tracking Control of sign & build of an R/C helicopter with minimum
Autonomous Robots.", Neural Processing Letters, pp.1-17, plane vibration. Then he received M.Sc. of Mec-
2015. hatronics from Tehran University in 2015. His
research interests include kinematics, dynamics
[30] Chen S, Yao B, Zhu X, Chen Z, Wang Q, Zhu S, Song
and control of robotic mechanical systems with
Y. "Adaptive Robust Backstepping Force Control of 1-DOF a particular emphasis on the Cable driven paral-
Joint Exoskeleton for Human Performance Augmentation.", lel mechanism. His works is based on applying new mathematical tools,
IFAC-PapersOnLine, No.480, Vol.19, 2016. such as convex optimization and heuristic algorithms to the end of pro-
viding real time algorithm for control and identification of this type of
[31] Liang, Yang, Liu Zhi, and Zhang Yun. "Adaptive fu-
robot. Also parts of his works include image processing and machine
zzy yaw moment control of humanoid robot based on ankle vision.
joint", 34th Chinese Control Conference (CCC), pp.5999-
6004. IEEE, 2014.
Received: 2016-05-09
Ali Amiri received the B.Sc. Degree in Electri- Accepted: 2017-01-10
cal Engineering from the K. N. Toosi University
of Technology Tehran, Iran, in 2011. Currently,
he is pursuing his Ph.D. from Tehran University.
His research interests include control of complex
system. His work is mainly on using different
control approaches like MPC in ROS.