2017 IEEE 30th Jubilee Neumann Colloquium • November 24-25, 2017 • Budapest, Hungary

Control Design-oriented qLPV Modelling of Virtual

Impedance-based Telemanipulation
József Kuti†∗ , Péter Galambos†

† Antal Bejczy Center for Intelligent Robotics, Óbuda University, Bécsi út 96/B. H-1034 Budapest, Hungary
∗ Doctoral School of Applied Informatics and Applied Mathematics, Óbuda University,

Bécsi út 96/B. H-1034 Budapest, Hungary

E-mail: [email protected], [email protected]

Abstract—The paper proposes a quasi-Linear Parameter Vary- Operator

ing (qLPV) model for the impedance model-based telemanipula- Fh
tion architecture with time-delay that relies only on measurable Master device
state variables and controllable considering practical design
purposes. This representation allows to formalize control design
criteria to handle properly the predictive and robustness aspects x
as well. The correctness of the derived model in terms of
approximation accuracy is proved via numerical simulations. F F b x
Index Terms—Telemanipulation, Time-delay, Force-feedback, m
Systems and Control, LPV/qLPV modelling

I. I NTRODUCTION 2 1
Slave device
The need for high quality force reflecting bilateral telema- Fs
nipulation raises in more and more field of robotics including Environment
medical devices, telesurgery, operating in remote or dangerous
environment, etc. One of the most critical problems is the com- Fig. 1. The impedance model based telemanipulation scheme with time-delay
munication delay between the master and the slave side. This
time-delay can decrease the transparency and the reliability of
break the transparency and the overall performance of the
the process and it can even destabilize it, see [1], [2], [3].
control system.
To handle the effect of time-delay, a lot of approaches were
The motivation of this paper is to model the system in such a
published during the last decades: the use of passive channel,
way that the predictive and the robust aspects can be taken into
its extension with passivity controller and observer [4], con-
account with a large variety of control criteria. Based on the
troller design based on the small-gain theorem [2], Lyapunov-
recent development of polytopic LPV model based methods
Razumikhin [5] or the Lyapunov-Krasovskii method [6] to
(see [19], [20]), the paper shows how a non-delayed, discrete
stabilize the system or by cutting the control loop by applying
time qLPV model can be derived for the virtual impedance
predictive controller [7] or sensory substitution [8].
model based telemanipulation scheme. This model allows for
To improve these concepts, plenty of approaches were designing controller that depends on the momentary degree of
published recently such as power based passivity control [9], time-delay and the estimated parameters of the environment
stabilizing controller design through approximating the system while it is robust to their uncertainties that are considered as
with non-delayed LPV system [10], [11], or by applying model parameters.
impulsive control [12]. The paper is structured as follows: Section II describes
Among the control architectures, the position-position, the impedance model based telemanipulation scheme and the
position-force, four channel structure, impedance model based modelling method. Then Section III shows the numerical
control and the predictive control scheme must be mentioned validation results and Section IV concludes the paper.
[13].
The predictive control-based approach can mitigate the II. D ERIVATION OF THE Q LPV MODEL
effect of the time-delay if the model of the remote environment The impedance model-based telemanipulation architecture
is known or it is well estimated. To achieve this, there are local [21] is considered along the paper that is depicted in Fig. 1.
estimator methods to approximate the actual properties like In this architecture, a damped virtual impedance model is
[14], and there are models for linear and nonlinear substances simulated based on the measured forces on the slave and the
[15], [16] and biological tissues [17], [18]. However, the master side. The communication time-delay denoted by 𝜏1 and
difference between the remote environment and its model can 𝜏2 is assumed to appear between the impedance model and

978-1-5386-4636-6/17/$31.00 ©2017 IEEE 000131

J. Kuti and P. Galambos • Control Design-oriented qLPV Modelling of Virtual Impedance-based Telemanipulation

the slave side. The critical time-delay that causes instability For discrete-time description, it can be written as
depends on the stiffness of the remote environment, the applied ˜ ⋅ (𝑥𝑠 [𝑖] − 𝑥𝑠 [𝑖 − 1]),
𝐹𝑠 [𝑖 + 1] = 𝐹𝑠 [𝑖] − 𝑘[𝑖] (6)
sampling time and the parameters of the impedance model [3].
This section investigates how it can be approximated in the ˜
where 𝑘[𝑖] = 𝑘𝑛 [𝑖]+Δ𝑘[𝑖] is the sum of a known nominal part
following typical LPV model form that is suitable for con- and an uncertain value. This uncertainty covers the difference
troller design combining the predictive and robust approaches: of the nominal 𝑘𝑛 value and the actual stiffness of the
⎡ ⎤ ⎡ ⎤ remote environment that is caused by the inaccuracy of the
x[𝑖 + 1] x[𝑖]
⎣ y[𝑖] ⎦ = S(p[𝑖]) ⎣ u[𝑖] ⎦ , applied stiffness model and the effect of time-delay in its
(1)
communication as well.
z[𝑖] w[𝑖]
The change of time-delay is crucial for quality of the
where the input signals are denoted by u[𝑖], the state variables telemanipulation because it can causes jumps in the signals.
by x[𝑖], the noises or disturbances by w[𝑖], the performance Although the effect of steps in the 𝐹𝑠 (𝑡) signal is mitigated
output z[𝑖], the measured output y[𝑖], and the S(p[𝑖]) system by the integrations, steps in 𝑥𝑠 (𝑡) value appear directly in the
matrix can be partitioned to A(p[𝑖]), B(p[𝑖]), C(p[𝑖]) etc. remote side.
standard matrices. Denote the time-delay of 𝑥𝑠 by 𝜏1 = Θ1 𝑇𝑠 and the time-
The model is called quasi-linear if the parameter- delay of 𝐹𝑠 by 𝜏2 = Θ2 𝑇𝑠 according to Figure 1. When
dependencies involves state variables to describe the nonlinear the connection to the slave device loses, the delay 𝜏1 starts
characteristics of the system. to increase and the slave-side does not receive new position
values so it suddenly stops. To avoid this phenomenon, a
A. Impedance model simple exponential slow-down is proposed here.
First of all, recall the continuous-time description of the Denote the dwelling time of the actual 𝑥𝑠 (𝑡) value by 𝜏1′ =
′
impedance model Θ1 𝑇𝑠 , that starts to increase from zero when the connection
loses. Instead of 𝑥𝑠 (𝑡) = 𝑥(𝑡 − 𝜏1 ), use
𝑚¨ ˙ + 𝐹∑ (𝑡),
𝑥(𝑡) = −𝑏𝑥(𝑡) (2) ′
˙ − 𝜏1 )𝑇 (1 − 𝑒−𝜏1 /𝑇 ),
𝑥𝑠 (𝑡) = 𝑥(𝑡 − 𝜏1 ) + 𝑥(𝑡 (7)
where 𝐹∑ (𝑡) stands for the sum of the measured force at the
master 𝐹ℎ (𝑡), at the slave device 𝐹𝑠 (𝑡), and the force denoted because in this case the velocity decreases continuously as
by 𝐹𝑐 (𝑡) used to stabilize the process and to mitigate the effect ′
˙ − 𝜏1 )𝑒−𝜏1 /𝑇 .
𝑥˙ 𝑠 (𝑡) = 𝑥(𝑡 (8)
of time-delay 𝜏 of the communication.
Taking into account the 𝑇𝑠 sampling time during the control, that disappears in 3𝑇 time. In discrete-time, it can be written
the following discrete-time state space model can be formal- as
ized for the impedance model by exploiting that the force 𝑇 ′ 𝑇𝑠
𝑥𝑠 [𝑖] = 𝑥[𝑖 − Θ1 ] + Δ𝑥[𝑖 − Θ1 ] (1 − 𝑒−Θ1 𝑇 ), (9)
values are measured with this sampling time as well: 𝑇𝑠
[ ] [ ][ ] [ ] where
𝑣[𝑖 + 1] 𝑎1 0 𝑣[𝑖] 𝑏
= + 1 𝐹∑ [𝑖], (3) Δ𝑥[𝑖] = 𝑥[𝑖] − 𝑥[𝑖 − 1]. (10)
𝑥[𝑖 + 1] 𝑎2 1 𝑥[𝑖] 𝑏2
where the constants are This way, its change can be written as
⎧

⎨ ∑ Δ𝑥[𝑖 − Θ1 ] 𝑖𝑓
ΔΘ1 = 0,
𝑎1 = 𝑒−𝑇𝑠 𝑏/𝑚 , 0
𝑚 Δ𝑥𝑠 [𝑖] = 𝑗=ΔΘ1 Δ𝑥[𝑖 − Θ1 + 𝑗] 𝑖𝑓
ΔΘ1 < 0,
𝑎2 = (1 − 𝑒−𝑇𝑠 𝑏/𝑚 ), 
⎩Δ𝑥[𝑖 − Θ ] 𝑇 (1 − 𝑒−Θ′1 𝑇𝑇𝑠 )
𝑏 1 𝑇𝑠 𝑖𝑓
ΔΘ1 = 1.
𝑏1 = (1 − 𝑒−𝑇𝑠 𝑏/𝑚 )/𝑏, (11)
( 𝑚) Furthermore, the large decrease of 𝜏1 can be avoided by
𝑏2 = 𝑇𝑠 − (1 − 𝑒−𝑇𝑠 𝑏/𝑚 ) /𝑏. temporally forbidding the use of the newest data.
𝑏
It is easy to see that assuming constant time-delays, the
The derived model is controllable and exactly characterizes
remote environment
the model in the sampling points.
𝐹𝑠 (𝑡 − 𝜏2 ) = 𝐹 (𝑥(𝑡 − 𝜏1 − 𝜏2 )) (12)
B. The slave-side
can be approximated as 𝜏˜1 = 𝜏1 +𝜏2 time-delay and 𝜏˜2 = 0. In
Now consider the remote environment
order to obtain a model with measurable state variables, this
𝐹𝑠 (𝑡) = 𝐹 (𝑥𝑠 (𝑡), 𝑥˙ 𝑠 (𝑡), 𝑥
¨𝑠 (𝑡)), (4) approximation will be applied for varying time-delays with the
previous auxiliary techniques as
where 𝑥𝑠 (𝑡) is the position of the slave device that is con-
trolled. Assume that it is passive and neglect its damping and ˜
𝐹𝑠 [𝑖 + 1] = 𝐹𝑠 [𝑖] − 𝑘[𝑖]Δ𝑥 𝑠 [𝑖] =
mass obtaining its model as a nonlinear spring ⎧
 ˜
𝐹𝑠 [𝑖] − 𝑘[𝑖]Δ𝑥[𝑖 − Θ] 𝑖𝑓 ΔΘ = 0,
⎨
𝑑𝐹𝑠 = −𝑘(𝑥𝑠 )𝑑𝑥𝑠 , (5) = ˜ ∑0
𝐹𝑠 [𝑖] − 𝑘[𝑖] Δ𝑥[𝑖 − Θ + 𝑗] 𝑖𝑓 ΔΘ < 0,
𝑗=ΔΘ

⎩𝐹 [𝑖] − 𝑘[𝑖]Δ𝑥[𝑖
˜ 𝑇 −Θ′1 𝑇𝑇𝑠
where 𝑘 denotes the actual stiffness. 𝑠 − Θ] 𝑇𝑠 (1 − 𝑒 ) 𝑖𝑓 ΔΘ = 1.

000132
 2017 IEEE 30th Jubilee Neumann Colloquium • November 24-25, 2017 • Budapest, Hungary

C. quasi–Linear Parameter Varying model 0𝑎×𝑏 denotes a zero matrix with size 𝑎 × 𝑏, the I𝑎×𝑏 matrix
Based on the former discrete-time equations, the parameter is the identity matrix with size 𝑎 × 𝑏 and the B matrix is the
varying model can be constructed. Its parameters are the same as in (15).
stiffness (here the nominal and the uncertainty), the sum of The state variables of the derived qLPV model are measur-
time-delay Θ and the ΔΘ value. able during the control process, furthermore, it is controllable
For sake of readability, first the constant Θ1 + Θ2 = Θ case for every 𝑘𝑛 + Δ𝑘 > 0. This way, it is stabilizable for all 𝜏
is considered. To handle this time-delay the state variables time-delay value.
must be chosen as
III. N UMERICAL VALIDATION
[
x[𝑖] = 𝑣[𝑖] Δ𝑥[𝑖] Δ𝑥[𝑖 − 1] . . . Δ𝑥[𝑖 − Θ] The numerical simulations proves that the described model
]𝑇
𝐹𝑠 [𝑖 − Θ] , (13) is exact
∙ if the Θ1 and Θ2 values are time invariant (see Fig. 2)
then the qLPV model
or
x[𝑖 + 1] = A(p)x[𝑖] + B ⋅ (𝐹𝑐 [𝑖] + 𝐹ℎ [𝑖]), (14) ∙ if Θ2 is zero (see Fig. 3).

where the system matrices are the followings

10-3
⎡ ⎤ 5
𝑎1 0 ... 0 0 𝑏1 ⎡ ⎤
⎢𝑎2 0 𝑏1 x0 (t)[m]
⎢ ... 0 0 𝑏2 ⎥⎥ ⎢𝑏2 ⎥
⎢0 1 ... 0 0 0⎥ ⎢ ⎥ x
qLPV
(t)[m]
⎢ ⎥ ⎢ ⎥
A(p) = ⎢ . .. .. .. ⎥ , B = ⎢ 0 ⎥ , 0
⎢ .. . . .⎥⎥ ⎢ .. ⎥
⎢ ⎣.⎦
⎣0 0 ... 1 0 0⎦
0
0 0 ... 0 −(𝑘𝑛 + Δ𝑘) 1 -5
(15) 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

that affinely depends on the 𝑘𝑛 and Δ𝑘 parameters. It is called

quasi-linear because the stiffness value can depend on position. 0.15
(t)[s]
1
This way, it describes a nonlinear characteristics in general. (t)[s]
Considering the time-varying delay, the model shows a 0.1 2

similar structure. By writing the time-delay as Θ1 + Θ2 = Θ

and assuming that its upper bound is known 0 ≤ Θ ≤ Θ𝑚 the 0.05

state variables must be chosen as

0
[ 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x[𝑖] = 𝑣[𝑖] Δ𝑥[𝑖] Δ𝑥[𝑖 − 1] . . . Δ𝑥[𝑖 − Θ𝑚 ] t[s]
]𝑇
𝐹𝑠 [𝑖 − Θ] , (16)
Fig. 2. Simulation results for constant 𝜏1 , 𝜏2 time-delays
and the A(p) system matrix of (14) qLPV model can be
written as In other cases, slowly increasing difference can be seen
⎡ ⎤ in the simulations, see the results in Fig. 4. This model
𝑎1 𝑏1
⎢ 𝑎2 0 2×(Θ𝑚 +1) ⎥ mismatch can be taken into account in an appropriately large
𝑏2
A(p) = ⎢ ⎥
⎣ 0Θ𝑚 ×1 IΘ𝑚 ×(Θ𝑚 +1) 0Θ𝑚 ×1 ⎦ , (17) Δ𝑘 uncertainty value.
The following parameters were used to derive the depicted
0 a3 (p) 1
results:
where if ΔΘ = 0 𝑇𝑠 5[𝑚𝑠] 𝑇 20[𝑚𝑠]
[ ] 𝑏 120[𝑁 𝑠/𝑚] 𝑚 1[𝑘𝑔]
a3 (p) = 01×Θ −(𝑘𝑛 + Δ𝑘) 01×(Θ𝑚 −Θ) ,
𝑘 1000[𝑁/𝑚] 𝐹ℎ (𝑡) sin(4𝑡)
if ΔΘ < 0
[ ] IV. C ONCLUSION
a3 (p) = 01×Θ 𝑎3 (p)11×(1−ΔΘ) 01×(Θ𝑚 −Θ+ΔΘ) ,
𝑎3 (p) = −(𝑘𝑛 + Δ𝑘), The paper showed the possibility and limitations of rewrit-
ing the impedance model based force reflecting telemanipu-
finally, if ΔΘ = 1 lation system with time-delay as a discrete-time non-delayed
[ ] qLPV model with measurable state variables.
a3 (p) = 01×Θ 𝑎3 (p) 01×(Θ𝑚 −Θ) , The numerical simulations showed that for constant time-
𝑇 ( ′ 𝑇𝑠
)
𝑎3 (p) = −(𝑘𝑛 + Δ𝑘) 1 − 𝑒−Θ1 𝑇 , delays the derived qLPV model is exact and it approximates
𝑇𝑠 well the dynamics even for varying time-delays.

000133
 J. Kuti and P. Galambos • Control Design-oriented qLPV Modelling of Virtual Impedance-based Telemanipulation

10-3 R EFERENCES
2
x0 (t)[m] [1] K. Gu and S.-I. Niculescu, “Survey on recent results in the stability and
1 control of time-delay systems,” Journal of dynamic systems, measure-
x (t)[m]
qLPV ment, and control, vol. 125, no. 2, pp. 158–165, 2003.
0 [2] K. Gu, J. Chen, and V. L. Kharitonov, Stability of time-delay systems.
Springer Science & Business Media, 2003.
-1 [3] P. Galambos, “Stability boundary of impedance controlled robots: effect
of stiffness, damping, friction and delay,” in Proceedings of the 15th
WSEAS International Conference on Systems. World Scientific and
-2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Engineering Academy and Society (WSEAS), 2011, pp. 247–252.
[4] J.-H. Ryu, D.-S. Kwon, and B. Hannaford, “Stable teleoperation with
time-domain passivity control,” IEEE Transactions on robotics and
0.2 automation, vol. 20, no. 2, pp. 365–373, 2004.
(t)[s] [5] M. Jankovic, “Control lyapunov-razumikhin functions and robust sta-
1
0.15 bilization of time delay systems,” IEEE Transactions on Automatic
(t)[s]
2 Control, vol. 46, no. 7, pp. 1048–1060, 2001.
0.1 [6] V. Kharitonov and A. Zhabko, “Lyapunov–krasovskii approach to the
robust stability analysis of time-delay systems,” Automatica, vol. 39,
0.05 no. 1, pp. 15–20, 2003.
[7] Z. Chen, Y.-J. Pan, and J. Gu, “Integrated adaptive robust control
0 for multilateral teleoperation systems under arbitrary time delays,”
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 International Journal of Robust and Nonlinear Control, vol. 26, no. 12,
t[s] pp. 2708–2728, 2016.
[8] M. J. Massimino and T. B. Sheridan, “Sensory substitution for force
feedback in teleoperation,” Presence: Teleoperators & Virtual Environ-
Fig. 3. Simulation results for varying 𝜏1 time-delay and 𝜏2 = 0 ments, vol. 2, no. 4, pp. 344–352, 1993.
[9] V. Chawda and M. K. OMalley, “Position Synchronization in Bi-
10-3 lateral Teleoperation Under Time-Varying Communication Delays,”
5 IEEE/ASME Transactions on Mechatronics, vol. 20, no. 1, pp. 245–253,
x0 (t)[m] Feb. 2015.
[10] P. Galambos and P. Baranyi, “TP𝜏 model transformation: A systematic
x (t)[m]
qLPV modelling framework to handle internal time delays in control systems,”
0 Asian Journal of Control, vol. 17, no. 2, pp. 486–496, 2015.
[11] J. Kuti, P. Galambos, and P. Baranyi, “Delay and stiffness dependent
polytopic lpv modelling of impedance controlled robot interaction,”
in Issues and Challenges of Intelligent Systems and Computational
-5
Intelligence. Springer International Publishing, 2014, pp. 163–174.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 [12] X. Li and S. Song, “Stabilization of delay systems: delay-dependent
impulsive control,” IEEE Transactions on Automatic Control, vol. 62,
no. 1, pp. 406–411, 2017.
0.2 [13] G. Niemeyer, C. Preusche, and G. Hirzinger, “Telerobotics,” in Springer
(t)[s] handbook of robotics. Springer, 2008, pp. 741–757.
1
0.15 [14] N. Diolaiti, C. Melchiorri, and S. Stramigioli, “Contact impedance
(t)[s]
2 estimation for robotic systems,” IEEE Transactions on Robotics, vol. 21,
0.1 no. 5, pp. 925–935, 2005.
[15] D. Erickson, M. Weber, and I. Sharf, “Contact stiffness and damping
0.05 estimation for robotic systems,” The International Journal of Robotics
Research, vol. 22, no. 1, pp. 41–57, 2003.
0 [16] G. Grioli and A. Bicchi, “A real-time parametric stiffness observer
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 for vsa devices,” in Robotics and Automation (ICRA), 2011 IEEE
t[s] International Conference on, 2011, pp. 5535–5540.
[17] Á. Takács, L. Kovács, I. Rudas, R.-E. Precup, and T. Haidegger, “Models
for force control in telesurgical robot systems,” Acta Polytechnica
Fig. 4. Simulation results for varying 𝜏1 , 𝜏2 time-delays Hungarica, vol. 12, no. 8, pp. 95–114, 2015.
[18] Á. Takács, P. Galambos, P. Pausits, I. J. Rudas, and T. Haidegger,
“Nonlinear soft tissue models and force control for medical cyber-
ACKNOWLEDGEMENTS physical systems,” in Systems, Man, and Cybernetics (SMC), 2015 IEEE
International Conference on, 2015, pp. 1520–1525.
Authors thankfully acknowledge the financial support of this [19] S. P. Boyd, Linear matrix inequalities in system and control theory.
work by the Hungarian State and the European Union under Philadelphia: Society for Industrial and Applied Mathematics, 1994.
[20] J. Kuti, P. Galambos, and P. Baranyi, “Polytopic TP Model based control
the EFOP-3.6.1-16-2016-00010 project and by the UNKP-16- analysis and synthesis concept,” in Proc. of the Int. Fed. of Aut. Contr.
3 and UNKP-16-4 New National Excellence Program of the (IFAC), 2017, pp. 6742–6747.
Ministry of Human Capacities and the support of the Doctoral [21] N. Hogan, “Impedance control: An approach to manipulation i-iii.”
Journal of dynamic systems, measurement, and control, vol. 107, pp.
School of Applied Informatics and Applied Mathematics of 1–24, 1985.
Óbuda University and Research and Innovation Center of
Óbuda University.

000134

View publication stats