Sei sulla pagina 1di 4

IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 18, NO.

4, AUGUST 2013

1431

Bilateral Teleoperation With Time-Varying Delay:


A Communication Channel Passification Approach
Yongqiang Ye, Ya-Jun Pan, and Trent Hilliard
AbstractIn this paper, the activeness of bilateral communication with
time-varying delay is analyzed. Based on the power-based time-domain
passivity control previously proposed, a time-domain passivity control approach is derived for bilateral communication. Teleoperation experimental
results verify the effectiveness.
Index TermsBilateral communication, teleoperation, time-domain
passivity control (TDPC).

Fig. 1.

Teleoperator with PTDPC.

A positive constant b is introduced to relate the different units of force


and velocity [10] and (1) can be written as follows:

I. INTRODUCTION

1 2
b 2
1
f (t) + vm
(t) (fm c bvm )2 (t)
2b m c
2
2b
b
1
1
+ fs2 (t) + vs2c (t) (fs + bvs c )2 (t)
2b
2
2b
1
1
= fm2 c (t) (fm c bvm )2 (t)
b
2b
1
2
+ bvs c (t) (fs + bvs c )2 (t)
2b
1
b 2
b
1
(t) vs2c (t).
+ fs2 (t) fm2 c (t) + vm
2b
2b
2
2

P =

Stability is always the first priority for teleoperation. The time delay, particularly time-varying delay, is a challenge for stability. Several
different solutions have been reported in the literature [1]. Latest teleoperation works addressing time-varying delay can be found in [2][5],
etc.
Remarkably, Hannaford and Ryu proposed the idea of time-domain
passivity control (TDPC) [6] and applied it to teleoperation with timevarying delay [7]. A passivity observer (PO) monitors the energy flow
into the system and when negative energy is observed, a time-varying
damping element called passivity controller (PC) is activated and dissipates the excessive energy. Since the PO monitors energy flow, the
approach is termed the energy-based TDPC (ETDPC). In [7], the communication is combined with the control software, termed as the bilateral controller, and ETDPC is applied to it. An alternative approach is
to apply ETDPC to the bilateral communication [8].
Recently, a simplified approach named power-based TDPC (PTDPC) has been proposed and applied to haptic interface control [9].
The note extends the application to teleoperation. For systems with
time-varying delay, if simple computation is required and transparency
is not important, one can choose this model-free architecture.

Using the transmissions, the last four items in (2) become


1
b 2
b
1 2
f (t) fm2 c (t) + vm
(t) vs2c (t)
2b s
2b
2
2
1
b 2
b 2
1 2
f (t) fs2 (t T2 ) + vm
(t) vm
(t T1 )
=
2b s
2b
2
2
 t
1 2
d
1
f ( )d T2 (t)fs2 (t T2 (t))
=
dt t T 2 2b s
2b
 t
b 2
b
d
2
v ( )d T1 (t)vm
(t T1 (t))
+
dt t T 1 2 m
2

II. TDPC FOR TELEOPERATION


In the bilateral communication, the forward velocity transmission is v s c (t) = vm (t T1 ) and the backward force transmission is
f m c (t) = fs (t T2 ), where T1 , 2 are the time delays (see Fig. 1).
The power flow equals the power entering from the master side
(vm (t)fm c (t)) minus the power exiting to the slave side (vs c (t)fs (t))
P = vm (t)fm c (t) vs c (t)fs (t).

(3)

where T1 (t) and T2 (t) are the changing rates of T1 (t) and T2 (t),
respectively, and (2) becomes
P =

(1)

Manuscript received May 12, 2012; revised August 31, 2012, October 30,
2012, and December 16, 2012; accepted March 16, 2013. Date of publication
April 15, 2013; date of current version July 8, 2013. Recommended by Technical Editor B. Shirinzadeh. This work was supported by the Natural Sciences
and Engineering Research Council of Canada and Canada Foundation for Innovation. The work of Y. Ye was supported by the Natural Science Foundation
of Jiangsu Province under Grant BK2010507 and Grant BK2012383, and by
the National Natural Science Foundation of China under Grant 61074161 and
Grant 61034005.
Y. Ye is with the College of Automation Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China (e-mail:
melvinye@nuaa.edu.cn).
Y.-J. Pan and T. Hilliard are with the Department of Mechanical Engineering, Dalhousie University, Halifax, NS B3H 4R2, Canada (e-mail: yajun.
pan@dal.ca).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TMECH.2013.2255882

(2)

1 2
1
fm c (t) (fm c bvm )2 (t)
b
2b
1
2
+ bvs c (t) (fs + bvs c )2 (t)
2b
b
1
2
(t T1 (t))
T2 (t)fs2 (t T2 (t)) T1 (t)vm
2b
2
dE
+
dt

where

E=
t T 2

1 2
f ( )d +
2b s

t
t T 1

b 2
v ( )d
2 m

(4)

(5)

can be viewed as stored energy.


Noting the passivity definition P = ddEt + Pd iss and the communication transmissions, the power dissipation of the bilateral communication is derived as
Pd iss =

1 2
1
1
f (t) (fm c bvm )2 (t) T2 (t)fm2 c (t)
b mc
2b
2b
1
b
+bvs2c (t) (fs + bvs c )2 (t) T1 (t)vs2c (t).
2b
2

1083-4435/$31.00 2013 IEEE

(6)

1432

IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 18, NO. 4, AUGUST 2013

TABLE I
PARAMETERS AND GAINS USED IN EXPERIMENTS

max T1,2
min T1,2

Value
0.5369s,0.542s
0.5s

Salve mass m
Ratio b

Value
0.1kg
2.5Ns/m

Slave P gain K
Slave D gain B

Value
370N/m
2.5Ns/m

max positive
max negative

d
T
dt 1,2
d
T
dt 1,2

Value
1
-34.7,-32.2

Pd iss 0 indicates passiveness and no PC action is needed. Note that


negative T1 , 2 (t) improves passiveness and additional energy is produced by the increasing delays.
Checking (6) in real time is not feasible because (6) involves signals
from both ports of the communication channel, which are not available
at the same time instant due to delay. To bypass the real-time checking
difficulty, checking (6) is separated into the checking of following two
terms sufficiently:

impedance is Fm c (s)/Vm (s) (uppercase letters represent the Laplace


transform of their lowercase letters) and the slave side impedance
is Fs (s)/Vs c (s) with ideal transparency. While with PCs, the master side impedance is changed to [Fm c (s) + FP C (s)]/Vm (s) and the
slave side impedance is Fs (s)/[Vs c (s) VP C (s)]. Note that in general [Fm c (s) + FP C (s)]/Vm (s) = Fs (s)/[Vs c (s) VP C (s)]. Thus,
the transparency is degraded by the output of PCs.

1
1
1 2
f (t) (fm c bvm )2 (t) T2 (t)fm2 c (t)
b mc
2b
2b
1
b
bvs2c (t) (fs + bvs c )2 (t) T1 (t)vs2c (t).
2b
2

III. EXPERIMENTAL VERIFICATION


(7)
(8)

Because (7) contains only the signals at the master side and (8) contains
only the signal at the slave side, checking (7) and (8) in real time is
straightforward. The price paid is the increased conservativeness as
in [7].
One can impose a positive upper bound of T1 , 2 (t), , as in [11] and
two POs are constructed at each side
1 2
1

f (t) (fm c bvm )2 (t) fm2 c (t)
b mc
2b
2b
1
b
= bvs2c (t) (fs + bvs c )2 (t) vs2c (t).
2b
2

Pomb sv =
Posb sv

(9)
(10)

Interestingly, if 2, Pomb sv and Posb sv are both nonpositive.


Two PCs attached at each port are activated when Pomb sv or Posb sv are
< 0 so that Pomb sv + Pcmtr = 0 or Posb sv + Pcstr = 0 where Pcmtr or Pcstr
is the power dissipation by the PCs. In real applications, zero division
must be avoided to prevent computation collapse.
Remark 1: The coefficient b is used for relating forces and velocities.
Similar to the wave impedance in [10], b presents a parameter which
can tune the intensity of damping. One can tune b as best suited to the
current task. Another design guideline is the principal of impedance
matching in wave-variable-based teleoperation.
In consideration of the signal flow directions in the communication
channel, the PC at the master side is of impedance causality and the
PC at the slave side is of admittance causality, as shown in Fig. 1.
Fig. 1 shows a teleoperator with PTDPC. The master and slave are
represented by effective endpoint masses of m. The master arm is
manipulated directly by an operator who exerts force fh . Since the
slave is controlled by a PD controller with PD gains K and B, the
slave is inherently passive. fs is the slave driving force generated by
the PD controller. fe is the environment force. The communication
channel transmits the velocity of the master vm for the slave to follow,
and it returns force feedback fs of the slave to the master. The actual
velocity of the slave is vs . The force perceived by the operator is
fm (t) = fs (t T2 ) + fP C (t) where fP C is the output of the master
side PC. The velocity command of the slave is vsn (t) = vm (t T1 )
vP C (t) where vP C is the output of the slave side PC. The PTDPC
makes the communication passive and if the environment is passive,
the teleoperator is passive.
This scheme not only increases conservativeness, but also trades
off transparency for passivity. If there is no PC, the master side

An in-house constructed interface system is used in the experiment.


The system is composed of a Quanser Q8 terminal board, a motor drive,
a shaft encoder, a brushless DC motor, and a steering wheel. The slave
and environment are virtual and rooted in a PC. The haptic feedback
force on the steering wheel is produced through a direct coupling to
an MCG IB34005 brushless DC motor with a peak torque of 5.2 Nm,
where the radius of the steering wheel is 16.9 cm. The motor is powered
by a BMC12L brushless servo drive, and the input signal is provided
by a Quanser Q8 data acquisition card mounted in the PC. The shaft
position is read with a 5000-line count shaft encoder. The sampling
rate is 1 kHz.
The master is the steering wheel. The slave is virtually constructed
in the PC as a mass of 0.1 kg. The design of b adopts the principal of
impedance matching, i.e., b equals the D gain of the slave PD controller.
The parameters used in the experiments are shown in Table I. The time
delays vary as 0.5 + rand s where rand indicates a uniform random
number between 0 and 0.1, but the changing rate of the delays is upperbounded by 1.
For clarity, the position/velocity signals at the master side and the
force signals at the slave side are shifted backward by the time delays
so that the signals at both sides can be viewed synchronously.
The operator turns the steering wheel so that the slave is driven to
contact with a virtual hard wall with stiffness K=30 kN/m located at
x = 0.6 m, as in [12]. The master position is defined as the rotational
distance of one point on the steering wheel.
When contact happens, the system exhibits strong oscillations without PTDPC. The operator can even hardly grip the steering wheel; see
Fig. 2(a). To verified the oscillation behavior without TDPC, simulation without PTDPC is also performed. In the simulation, the model
of the human operator in [13] is adopted where the human operator is
modeled as a PD-type position tracking controller with its spring and
damper gains as 75 N/m and 50 Ns/m, respectively. The simulation
again confirm the instability; see Fig. 2(b).
With the help of PTDPC, steady contact is achieved. The master and
slave positions are shown in Fig. 3(a), indicating that the contact happens at 2.8 s. The fluctuation of the slave position after the hard contact
point is trivial. The environment force fe is recorded in Fig. 3(b), where
the first force spike at the contact point is 20 N and the following forces
decay within 2 s. Fig. 3(c) and (d) shows the PCs output fP C (with
largest one 2 N) and vP C (with largest one 2.2 m/s); both vanish
after the fifth second. The forces across the communication channel
fm and fs are shown in Fig. 3(e). The most noticeable spikes of fm
and fs are both 4.6 N. The discrepancy between fm and fs after the
contact point is insignificant. Fig. 3(f) shows the time delays T1 , 2 .

IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 18, NO. 4, AUGUST 2013

Fig. 2.

1433

Master and slave positions without PTDPC. (a) Experiment. (b) Simulation.

Fig. 4(a) and (b) shows the two PO values. The master side PO (Pm )
fluctuates and the slave side PO (Ps ) is mostly negative. The overall
dissipation Pd iss (Pd iss = Pm + Ps ) in Fig. 4(c) is quite similar to
Ps , which means that Pd iss is dominated by Ps . The energies across
the communication channel after the PCs compensation, i.e., the total
energy Et and the energies at the master side and slave side, Em and
Es (Et = Em + Es ), all increase first and keep at positive values in
Fig. 4(d). This proves the effective working of the PCs.
IV. CONCLUSION AND FUTURE WORK
Bilateral communication with time delay can lead to instability in
teleoperation. A PTDPC scheme is proposed for dealing with the instability. Two sufficient conditions replace the original passivity condition
that is hard to be checked due to time delay. Two POs and two PCs
are attached at each port of the communication channel. The PCs make
each port passive and in turn stabilize the teleoperator. The verification test with asymmetric and time-varying delays is successful. This
model-free approach can be adopted when simple computation is required and transparency is not important. Future work will investigate
more realistic application.
Fig. 3. PTDPC for bilateral teleoperation 1. (a) Master and slave positions.
(b) Contact force fe . (c) Master side PC output fP C . (d) Slave side PC output
v P C . (e) fm and fs . (f) T 1 and T 2 .

ACKNOWLEDGMENT
The work of Y. Ye was initiated in the ACM-Lab, Dalhousie University. The authors would like to thank the reviewers and the Associate
Editor for their valuable comments that have improved the quality of
this paper.
REFERENCES

Fig. 4. PTDPC for bilateral teleoperation 2. (a) Master side PO value P m .


(b) Slave side PO value P s . (c) Sum of two PO values (P d iss ). (d) Energy values
(E t , E m , and E s ).

[1] P. F. Hokayem and M. W. Spong, Bilateral teleoperation: An historical


survey, Automatica, vol. 42, no. 12, pp. 20352057, Dec. 2006.
[2] S. Munir and W. J. Book, Internet-based teleoperation using wave variables with prediction, IEEE/ASME Trans. Mechatronics, vol. 7, no. 2,
pp. 124133, Jun. 2002.
[3] K. Walker, Y.-J. Pan, and J. Gu, Bilateral teleoperation over networks
based on stochastic switching approach, IEEE/ASME Trans. Mechatronics, vol. 14, no. 1, pp. 539554, Oct. 2009.
[4] C.-C. Hua and X. P. Liu, A new coordinated slave torque feedback control
algorithm for network-based teleoperation systems, IEEE/ASME Trans.
Mechatronics, vol. 18, no. 2, pp. 764774, Apr. 2013.
[5] E. Delgado, M. Diaz-Cacho, D. Bustelo, and A. Barreiro, Generic approach to stability under time-varying delay in teleoperation: Application to the position-error control of a gantry crane, IEEE/ASME Trans.
Mechatronics, to be published.
[6] B. Hannaford and J.-H. Ryu, Time-domain passivity control of haptic
interfaces, IEEE Trans. Robot. Autom., vol. 18, no. 1, pp. 426434, Feb.
2002.

1434

[7] J.-H. Ryu and C. Preusche, Stable bilateral control of teleoperator under
time-varying communication delay: Time domain passivity approach,
in Proc. IEEE Int. Conf. Robot. Autom., Roma, Italy, 2007, pp. 3508
3513.
[8] A. Iqbal and H. Roth, Stabilization of teleoperated systems with stochastic time delays using time domain passivity control, in Proc. SICE-ICASE
Int. Joint Conf., Busan, Korea, 2006, pp. 393398.
[9] Y. Ye, Y.-J. Pan, Y. Gupta, and J. Ware, A power-based time domain
passivity control for haptic interfaces, IEEE Trans. Control Syst. Technol.,
vol. 19, no. 4, pp. 874883, Jul. 2011.

IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 18, NO. 4, AUGUST 2013

[10] G. Niemeyer and J.-J. E. Slotine, Telemanipulation with time delays,


Int. J. Robot. Res., vol. 23, no. 9, pp. 873890, Sep. 2004.
[11] R. Lozano, N. Chopra, and M. W. Spong, Passivation of force reflecting
bilateral teleoperators with time varying delay, presented at the Mechatronics 2002, Enschede, The Netherlands, Jun. 2002.
[12] M. Mehrtash, N. Tsuda, and M. Khamesee, Bilateral macro-micro teleoperation using magnetic levitation, IEEE/ASME Trans. Mechatronics,
vol. 16, no. 3, pp. 459469, Jun. 2011.
[13] D. Lee and M. W. Spong, Passive bilateral control of teleoperators under
constant time delay, IEEE Trans. Robot., vol. 22, no. 2, pp. 269281,
Apr. 2006.

Potrebbero piacerti anche