You are on page 1of 8

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/331716114

Stabilization of a Two-DOF Spherical Parallel Robot via a Novel Adaptive


Approach

Conference Paper · March 2019


DOI: 10.1109/ICRoM.2018.8657566

CITATIONS READS
6 213

6 authors, including:

Saeed Ansari-Rad Mojtaba Zarei


University of Tehran Duke University
11 PUBLICATIONS   67 CITATIONS    28 PUBLICATIONS   147 CITATIONS   

SEE PROFILE SEE PROFILE

Mehran Ghafarian Tamizi Mehdi Tale Masouleh


Mehdi Tale Masouleh's Lab University of Tehran, School of Electrical and Computer Engineering
3 PUBLICATIONS   9 CITATIONS    196 PUBLICATIONS   1,414 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Control of a 6-DoF pneumatically actuated Gough-Stewart platform View project

ShrewdShoe View project

All content following this page was uploaded by Mojtaba Zarei on 29 May 2019.

The user has requested enhancement of the downloaded file.


Stabilization of a Two-DOF Spherical Parallel
Robot via a Novel Adaptive Approach
Saeed Ansari-Rad Mojtaba Zarei Mehran Ghafarian Tamizi
School of Electrical and Computer Electrical and Computer Engineering School of Electrical and Computer
University of Tehran Duke University University of Tehran
Tehran, Iran NC, USA Tehran, Iran
Saeedansari71@ut.ac.ir mojtaba.zarei@duke.edu mehran.ghafarian@ut.ac.ir

Saeid Mohammadi Nejati Mehdi Tale Masouleh Ahmad Kalhor


School of Electrical and Computer School of Electrical and Computer School of Electrical and Computer
University of Tehran University of Tehran University of Tehran
Tehran, Iran Tehran, Iran Tehran, Iran
saeid.nejati.moh@ut.ac.ir m.t.masouleh@ut.ac.ir a.kalhor@ut.ac.ir

Abstract—This paper proposes a control strategy for using a In order to promote the performance of stabilizer, a proper
2-Degree-of-Freedom (DOF) parallel robot as a camera stabilizer. setup should be designed. This setup contains a lightweight,
To configure the stabilizer, a gyro sensor is mounted on the swift and precise manipulator such as those introduced in [7]
end-effector of the robot and the data is transmitted to a
PC in order to permanently command two servo motors to and [8]. A stabilizer mechanism is developed in [9–11], by
the end of compensating the external disturbances. A novel using spherical ultrasonic and shape-memory alloy actuators.
adaptive approach is utilized to stabilize the end-effector of the Surveying through diverse setups, parallel manipulators can be
2-DOF Spherical Parallel Robot under study, where an adaptive regarded as the proper suggestion for stabilization purpose due
parameter should be adjusted in order to damp disturbances, to their high accuracy and speed, such as Spherical Parallel
exponentially. To assess the method, first in SimMechanics,
operation of the 2-DOF robot with the adaptive approach is Robots (SPRs), which only perform rotational movement.
simulated and examined. In addition, a well-known and robust Hesar et al. [12] have suggested an image-based control
decay algorithm is employed to show the superiority of the strategy for a 2-DOF SPR to follow moving objects. For the
proposed method. Then, by implementing on the real structure, same plant, Bozorgi et al. [13] prepared dynamic analysis and
the performance of the adaptive controller is validated. The a torque-based method which makes the end-effector to follow
proposed method leads to better performance in compared with
the decay algorithm, from view point of quantitative indices of the desired trajectories. The mentioned studies show high
tracking, stabilization and smoothness. velocity, stiffness and dynamic performance of SPRs; they
Keywords—2-DOF parallel robot, Stabilization, Adaptive con- are applicable in industry, where performing tasks in limited
trol, Decay exponentially algorithm spaces is of importance [14]. The under-study 2-DOF SPR,
named as TezGoz, is one of the spherical robots designed in
I. I NTRODUCTION the Human and Robot Interaction Laboratory at the University
of Tehran.
Image capturing is one of important applications of mobile In order to configure the stabilizer, a gyro sensor is often
robots, which is utilized for gathering data from unknown mounted on the dynamic platform [15]. The optimal method
environments. A main challenge of autonomous robots is how proposed in [16] consists in applying an indirect strategy,
to take images with high quality. In this way, image stabilizers where the results of this method are also compared with a lead-
are vital part of these robots and as a result, designing an PI controller. Tunning the constant controller’s parameters to
agile setup is a prerequisite. In the literature, image stabilizing damp the system’s oscillation has been investigated in several
methods, generally, fall into four groups. The first group has papers. In [17, 18], oscillation number index; which is calcu-
been constructed on hardware-based methods and utilizes me- lated based on the arc-length based Lyapunov functions [19],
chanical devices to keep the observable pathway of camera [1]. was utilized to alleviate the system’s oscillation. The aforesaid
The second approach, referred to as optical image stabilization, indices dealt with the modeled systems with the limited
relies on controlling an image sensor or lens to keep the optical interval of uncertainties which are hardly applicable to the
path at the position on image sensor [2, 3]. The third method stabilization of the agile systems with correlative uncertainties
known as electronic image stabilization, uses an incorporated in their structures. Among these varied control strategies,
inertial sensor [4] and finally the fourth method that is called adaptive controllers [20, 21], in despite of their adaptability
digital image stabilization, uses a moving vector computed to circumstance conditions, were disregarded.
from feature matching or tracking algorithms [5, 6]. In this paper, via a novel adaptive controller, the end-
{E} coordinate. To control the position on desired trajectories
and stabilize the end-effector, actuators positions should be
held on specific desired value, which can be calculated using
the following formulations:[22]:
B ∗
ηy
θ1∗ = tan−1 (− B
) (4)
ηz∗
B ∗B ∗
η η
θ2∗ = tan−1 (− B ∗ 2 x Bz ∗ 2 ) (5)
ηy + ηz
where B ηj∗ , j = x, y, z is the desired value of end-effector
orientation along the j-axis, described in {B}. These pa-
rameters can be calculated by applying the rotation matrix
B B W W ∗
Fig. 1. Mechanical structure of TezGoz [22]. W R = E R(E R) to
|
η , where B W R denotes the rotation
matrix describing {W} relative to {B}. Also, it is required for
effector of a 2-DOF SPR is stabilized. Using SimMechanics angular velocity of actuators (θ˙i , i = 1, 2) to be held on the
simulations and implementation on the real plant, the perfor- following desire values[22]:
mance of the adaptive controller is compared with a decay
 ∗   
θ̇1 θ̇1
algorithm [22] as a robust and well-known approach. The = − J I (B E
E R ω E\W ) (6)
θ̇2∗ θ̇2
remainder of paper is organized as follows. After describing
the robot structure and kinematic equations in Section II, the Substituting (1) into (6) and calculating B
E R, one has:
adaptive approach and its corresponding theory are explained  ∗   
in Section III. Simulation results are included in Section IV, θ̇1 θ̇1
= + J E ω E\W (7)
and Section V is devoted to implementation of the proposed θ̇2∗ θ̇2
algorithm on the real plant to validate the results. where:
II. K INEMATIC AND S TABILIZATION A NALYSES OF THE  − cos θ1 cos θ2 sin θ2 
3 0 3
2-DOF SPR (1−sin2 θ 1 cos2 θ 2) 2 (1−sin2 θ 1 cos2 θ 2) 2
J = 1
(1−sin2 θ1 cos2 θ2 ) 2 − cos θ2 tan θ1

In this part, the mechanical structure of the TezGoz, the 2- 0 cos θ1 1
(1−sin2 θ 1 cos2 θ 2) 2
DOF SPR under study in this paper, consists of 2 servo motors, (8)
3 links and the end-effector. The kinematic arrangement is It is necessary to mention that E ω E\W and W
E R could be
such that the end-effector is connected to one of the servos obtained from the IMU sensor mounted on the end-effector.
by a link (In the figure, denoted as Link 1) and connected
to another servo by 2 links (Link 2 and Link 3). All axes III. A DAPTIVE C ONTROL D ESIGN T HEORY
have intersection at a common and certain point which is The main objective of the controller is to track a nonlinear
called wrist point. It is worth mentioning that the axes of desirable path of the actuator angle and angular velocity,
any 2 adjacent joints are perpendicular to each other. In the which have been presented in Section II, while compensating
following, the frame attached to the body of robot is {B} and the system’s uncertainties. The corresponding error states are
the end-effector frame is shown with {E}. More details of presented by the following relations:
robot structure are discussed at [14, 22].
In [14], the kinematic model of TezGoz is completely ex- 
ei = θi − θi ∗
plained. Using architectural parameters of TezGoz, the inverse ; i = 1, 2 (9)
ėi = θ̇i − θ̇i∗
Jacobian matrix of robot is obtained:
In the above equations (9), θi∗ and θ̇i∗ , i = 1, 2 are denoted
 
1 0 0
JI =
− tan θ1 cos θ2 sin θ2 −1 − tan θ1 sin2 θ2 to the desired angle and velocity of the actuators, respectively.
(1) In the case of parameter certainty, the governing equation on
where θi , i = 1, 2 is the position of actuators positions. To the error is calculated to damp the errors:
stabilize the robot, the end-effector orientation (η) must be
fixed with respect to the ground frame {W}. This goal can be
formulated as follows: ėi = −κei , i = 1, 2. (10)
W
η ∗ = cte (2) Moreover, the system’s inputs (u) are the velocity of the
actuators which can be presented by θ̇ = u. Due to the fact
W
ω ∗E/W =0 (3)
that in the experimental setup, the actuator’s reactions probably
where W η ∗ is the desired value of the end-effector orientation have some degree of interaction on each other, one can define
relative to {W} and W ω ∗E/W is the desired value of the end- a correlative representation of the aforementioned formulation
effector angular velocity relative to the ground described in the as below:
Employing some mathematical calculations, one can calcu-
      late the adaption rules as follows:
θ̇1 u β β2
=B 1 , B= 1 (11) e1 e2
θ̇2 u2 β3 β4 β̇1 = u1 , β̇2 = u1 , β̇3 = eγ1 u2 , β̇4 = eγ2 u2
γ γ
where βi , i = 1, · · · , 4 stands for the unknown parameters eu|
→ Ḃ = (21)
in the system which should be adjusted in order to damp the γ
system’s error, exponentially. Having in mind the aforesaid It is obvious that Matrix (B) is full rank in the ideal
formulation, one can define the following Lyapunov function: conditions. On the other hand, in the case of uncertainty, by
employing relatively significant values for γ, matrix B re-
4 mains diagonally dominant in (17), and as the result invertible.
1 γ X e2
V = ||e||2 + β (12)
2 2 i=1 i IV. S IMULATION S TUDY
e = B − Bd
B (13) In this part, based on the analytical formulas, a SimMe-
chanics model of 2-DOF spherical robot is developed by
where in the above, B e stands for the difference between the using MATLAB software. In Fig. 4, this simulation model
uncertain input matrix (B) and the actual value of the input is illustrated. The adaptive approach or stabilization is im-
matrix (B d ). Moreover, γ is an arbitrary sufficient big positive plemented in the controller block; other elements such as
scalar value which determines the sensitivity amount of the sensors, the disturbance table and the plant are also shown.
adaptation rules to the changes. The derivative of the proposed By applying external excitation to the disturbance table, co-
Lyapnov function leads to: ordinates of the base axes are changed. The incentive of this
paper is to stabilize end-effector orientations, while bearing
4 these excitation signals. In this section, the effect of time-
X ˙
V̇ = ė| e + γ βei βei (14) varying and sinusoidal excitations is studied, where in compare
i=1 to other types of signals, they are hard ro track. To show
Considering (9), (11) and (13), (14) is reformulated as performance of the control scheme while undertaking different
follows: uncertainty and errors, external disturbances and gyro noises
are considered in the simulation.
4 Three important indices
are defined for quantitative analysis:
∗ |

Jstable = W η − W η ∗ is utilized to show the orientation of
X
V̇ = (B d u − θ̇ ) e + γ β̇i βei (15)
i=1 end-effector toward the ground frame from the desired value,
therefore the stabilization
purpose can be surveyed in each
Employing some mathematical calculations, the first term
moment. Jtrack = 0.01 θ shows the precision of tracking
e
of the right side of (11) is reformulated as bellow:
the desired trajectories, where θe = θ − θ∗ .
∗ | ∗
Jsmooth = 104 ku̇k is the third index, which shows smooth-
e + Bu − θ̇ )| e
(B d u − θ̇ ± Bu) e = (−B (16) ness of input signals and is defined due to limitations of
actuators in implementation.
In order to hold the proposed condition on the errors, one These indices show the performance of control methods

can consider Bu − θ̇ = −Ke. Thus, the control signal is applied to the 2-DOF robot; whenever the mentioned values
obtained by the following equation: cross the acceptable value of each index, the applied method
∗ imposes some difficult condition to the plant. The external
u = B −1 (θ̇ − Ke) (17) disturbance on actuators and noises on the gyro sensors are
shown with xdist and xnoise , respectively. In the simulations,

external disturbances are set as constant signals (xdist = 0.5 )
θ − θ1 ∗
     ∗
θ̇1u θ̇
= B −1 (−K 1 + 1∗ ) (18) and the gyro noises are considered as white banded noise
θ̇2u θ2 − θ2 ∗ θ̇2
with power of 0.01. Another realistic condition is devoted
Thus, the derivative of the proposed Lyapunov function is to saturation of control effort signals, which are limited to
reformulated as below: 2(rad/sec) for each actuator. Dynamixel MX-64 as our target
motor for implementation faces this condition. The harmonic
4
X of external excitations around each axis is chosen as
V̇ = −Ke| e − u| Be + γ β̇i βei (19)
i=1 θexcj = aj sin(fj t)(rad), j = x, y, z (22)
Expanding (19) yields: If aexc and fexc are defined such that aexc = ax = ay = az
and fexc = fx = fy = fz , then accepted values of each indices
V̇ = −Ke| e − (u1 β1 + u2 β3 )e1 − (u1 β2 + u2 β4 )e2 for a control strategy can be stated based on these parameters.
+γ β̇1 βe1 + γ β̇2 βe2 + β̇3 βe3 + γ β̇4 βe4 (20) In this paper, by selecting aexc = 0.4(rad), accepted values
Base TABLE I
Orientation(degree) 20
End-effector 2 C OMPARING THE L2 - NORM OF INDICES FROM SIMULATIONS .

10
0 Method kJstable kL2 kJtrack kL2 kJsmooth kL2
Adaptive 1.5101 2.6376 0.2729
-2
0 Decay 1.7013 2.9684 0.2839
0 5 10 0 5 10
Time(sec) Time(sec)
4
2
IMU Noise

2
TABLE II
0 0 C OMPARING THE L2 - NORM OF INDICES FROM IMPLEMENTATION .
-2
-2
-4 Method kJstable kL2 kJtrack kL2 kJsmooth kL2
0 5 10 0 5 10
Time(sec) Time(sec)
Adaptive 1.7894 2.8375 0.1281
Decay 1.9351 3.1293 0.1317
Fig. 2. Adaptive control algorithm simulation results.

For example, with the adaptive controller, despite nearly se-


of Jstable , Jtrack and Jsmooth for adaptive approach are set
vere disturbance signals, the orientation value ( W η − W η ∗ )
in frequency of fexc = 1.5(rad/sec). These accepted values
never crosses 1 degree, while the decay algorithm with same
demonstrate whether behaviour of stabilizer in term of indices
conditions crosses 1 degree several times during the simula-
is appropriate or not; in Fig. 3, these accepted values are plot-
tion.
ted for under-study stabilizers. For other types of excitation,
Table I compares the two control methods from the L2 -
these acceptable values are still definable.
norm of indices. Despite Fig. 3, where peaks of indices are of
importance, in Table 1 the average value of parameters is uti-
1
J stable(degree)

lized as criterion. Likewise, from the foregoing results, it can


0.8
0.6
be inferred that the adaptive scheme has better performance
0.4
rather than the decay algorithm.
The selected values for two schemes are optimal. For ex-
0 1 2 3 4 5 6 7 8 9 10
ample, by increasing the gain of decay algorithm, the index of
2 tracking increases; but the smoothness index of control efforts
J track (rad )

exceeds from accepted value. Therefore, actuators encounter


1
harsh variation of input signals, which is not desirable and
0 may pose significant damages to motors.
1 2 3 4 5 6 7 8 9 10
4
V. I MPLEMENTING ON 2-DOF S PHERICAL PARALLEL
J smooth (rad/S2 )

ROBOT
2
In this section, the adaptive approach is implemented on
0 the 2-DOF parallel robot under study; setup components are
0 1 2 3 4 5 6 7 8 9 10
shown in Fig.6. In this structure, two Dynamixel Mx-64
Time(s)
servo motors are utilized. Moreover, a 9-axis IMU sensor
Fig. 3. Comparing indices of the two methods with accepted values. named MPU9150 is attached to the end-effector to measure
orientations and angular velocities. To process these informa-
Thereafter, in order to verify advantages of the proposed tion, an Arduino board is the link between sensors and PC.
approach, an exponentially decay algorithm is applied in The PC commands the servos expansion board that provides
simulation examples. In these simulations, amplitude of base either Dynamixel output of RS232 or RS485; Usb2dynamixel
excitation signals are set to ax = ay = az = 0.3(rad) and transmits data from PC to servos.
az = 0.1(rad), also frequencies of these signals are set as Results of the adaptive approach implementation are demon-
fx = 6.5(rad/sec), fy = 4.5(rad/sec) and fz = 2.5(rad/sec). strated in Fig.5; this method succeeds to stabilize base ex-
The parameters of adaptive controller are chosen as γ = citations (these excitations are applied to the base by hand
104 , Kadaptive = 300 and the gain of decay algorithm is set motion). To compare with the mentioned decay algorithm,
Kdecay = 200, as well. indices of tracking, stabilization and smoothness (from Section
Results of simulations are demonstrated in Fig. 2, where IV) for implementation of both methods, are reported in Table
the adaptive control scheme succeeds to stabilize the objective II and show the superiority of the proposed adaptive method.
end-effector in existence of considerable uncertainties and
measurement noises. C ONCLUSION
In Fig. 3, the indices are compared for both methods. From In this paper, a novel adaptive approach for stabilization of a
stability and tracking points, both Jstable and Jtrack show that 2-DOF parallel robot was proposed. By acquiring desired posi-
the proposed adaptive scheme leads to a better performance. tion and angular velocities of actuators from inverse kinematic,
Fig. 4. SimMechanics block diagram.

35
robot and succeeded to damp significant excitations. As ongo-
Base ing work, the dynamic model of the robot will be obtained
30 End-effector
for controlling the robot; machine learning approaches are
25 powerful tools for this purpose.
Orientation (degree)

20
R EFERENCES
15
[1] M. k. Masten, “Inertially stabilized platforms for optical
10 imaging systems,” IEEE Control Systems, vol. 28, no. 1,
5 pp. 47–64, 2008.
0
[2] J. Egbert and R. W. Beard, “Low-altitude road following
0 2 4 6 8 10
Time (s)
12 14 16 18 20
using strap-down cameras on miniature air vehicles,”
Mechatronics, vol. 21, no. 5, pp. 831–843, 2011.
Fig. 5. Adaptive control algorithm implementation results. [3] D. Sachs, S. Nasiri, and D. Goehl, “Image stabilization
technology overview,” InvenSense Whitepaper, 2006.
[4] A. Amanatiadis, A. Gasteratos, S. Papadakis, and
the controller approach is utilized to track these desired values. V. Kaburlasos, Image stabilization in active robot vision.
Consequently, the orientation of the end-effector of robot INTECH Open Access Publisher, 2010.
remains invariable. The adaptive parameter of controller varies [5] Y. Qiang, W. Baifeng, J. Yi, and Z. Kun, “Summarization
due to the fact that reactions of actuators have some degree of of electronic image stabilization,” in 2006 7th Interna-
interaction on each other and the adaptive approach succeed to tional Conference on Computer-Aided Industrial Design
damp disturbances of system. Simulating the 2-DOF robot in and Conceptual Design. IEEE, 2006, pp. 1–5.
SimMechanics, indices of tracking, stabilizing and smoothness [6] T. Van Lanh, K.-S. Chong, S. Emmanuel, and M. S.
demonstrated advantages of the adaptive approach opposing Kankanhalli, “A survey on digital camera image forensic
with a decay exponentially algorithm. Despite the decay methods,” in 2007 IEEE international conference on
method that is quite similar to proportional controllers and multimedia and expo. IEEE, 2007, pp. 16–19.
shows robust and vigorous results, the adaptive approach acted [7] D.-Y. Koh, Y. K. Kim, K.-S. Kim, and S. Kim,
superior from the point of mentioned indices. Thereafter, the “Bioinspired image stabilization control using the adap-
adaptive method was implemented on a real 2-DOF parallel tive gain adjustment scheme of vestibulo-ocular reflex,”
USB to Dynamixel
Expansion board

MPU 9150 Arduino board

Fig. 6. Test setup of 2-DOF parallel robot

IEEE/ASME Transactions on Mechatronics, vol. 21, Second RSI/ISM International Conference on. IEEE,
no. 2, pp. 922–930, 2016. 2014, pp. 445–450.
[8] H. Li, J. Luo, C. Huang, Q. Huang, and S. Xie, “Design [14] A. Arian, B. Danaei, and M. T. Masouleh, “Kinematics
and control of 3-dof spherical parallel mechanism robot and dynamics analysis of a 2-dof spherical parallel
eyes inspired by the binocular vestibule-ocular reflex,” robot,” in Robotics and Mechatronics (ICROM), 2016 4th
Journal of Intelligent & Robotic Systems, vol. 78, no. International Conference on. IEEE, 2016, pp. 154–159.
3-4, pp. 425–441, 2015. [15] P. J. Kennedy and R. L. Kennedy, “Direct versus indirect
[9] M. Hoshina, T. Mashimo, and S. Toyama, “Development line of sight (los) stabilization,” IEEE Transactions on
of spherical ultrasonic motor as a camera actuator for control systems technology, vol. 11, no. 1, pp. 3–15,
pipe inspection robot,” in 2009 IEEE/RSJ International 2003.
Conference on Intelligent Robots and Systems. IEEE, [16] A. A. Roshdy, Y. Z. Lin, C. Su, H. F. Mokbel, and
2009, pp. 2379–2384. T. Wang, “Design and performance of non-linear fuzzy
[10] J.-s. Lee, D.-k. Kim, S.-w. Baek, S.-h. Rhyu, and B.-i. logic pi controller for line of sight stabilized platform,”
Kwon, “Newly structured double excited two-degree-of- in Optoelectronics and Microelectronics (ICOM), 2012
freedom motor for security camera,” IEEE Transactions International Conference on. IEEE, 2012, pp. 359–363.
on Magnetics, vol. 44, no. 11, pp. 4041–4044, 2008. [17] M. Zarei, A. Aflakian, A. Kalhor, and M. T. Masouleh,
[11] T. Wolfe, M. Faulkner, and J. Wolfaardt, “Development “Oscillation damping of nonlinear control systems based
of a shape memory alloy actuator for a robotic eye on the phase trajectory length concept: An experimental
prosthesis,” Smart materials and structures, vol. 14, case study on a cable-driven parallel robot,” Mechanism
no. 4, p. 759, 2005. and Machine Theory, vol. 126, pp. 377–396, 2018.
[12] M. E. Hesar, M. Masouleh, A. Kalhor, M. Menhaj, and [18] M. Zarei, A. Kalhor, and M. Rastegar, “Employing phase
N. Kashi, “Ball tracking with a 2-dof spherical parallel trajectory length concept as performance index in lin-
robot based on visual servoing controllers,” in Robotics ear power oscillation damping controllers,” International
and Mechatronics (ICRoM), 2014 Second RSI/ISM Inter- Journal of Electrical Power & Energy Systems, vol. 98,
national Conference on. IEEE, 2014, pp. 292–297. pp. 442–454, 2018.
[13] E. R. J. Bozorgi, I. Yahyapour, A. Karimi, M. T. Ma- [19] M. Zarei, A. Kalhor, and D. Brake, “Arc length based
souleh, and M. Yazdani, “Design, development, dynamic maximal lyapunov functions and domains of attraction
analysis and control of a 2-dof spherical parallel mech- estimation for polynomial nonlinear systems,” Automat-
anism,” in Robotics and Mechatronics (ICRoM), 2014 ica, vol. 90, pp. 164–171, 2018.
[20] S. Ansari-Rad, A. Kalhor, and B. N. Araabi, “Partial
identification and control of mimo systems via switching
linear reduced-order models under weak stimulations,”
Evolving Systems, pp. 1–18, 2017.
[21] S. Ansari-Rad, S. Jahandari, A. Kalhor, and B. N. Araabi,
“Identification and control of mimo linear systems under
sufficient and insufficient excitation,” in 2018 Annual
American Control Conference (ACC). IEEE, 2018, pp.
1108–1113.
[22] B. Danaei, M. Alipour, A. Arian, M. T. Masouleh,
and A. Kalhor, “Control of a two degree-of-freedom
parallel robot as a stabilization platform,” in 2017 5th RSI
International Conference on Robotics and Mechatronics
(ICRoM). IEEE, 2017, pp. 232–238.

View publication stats

You might also like