This paper presents a novel teleoperation controller for a nonlinear master–slave robotic system with constant time delay in communication channel. The proposed controller enables the teleoperation system to compensate human and environmental disturbances, while achieving master and slave position coordination in both free motion and contact situation. The current work basically extends the passivity based architecture upon the earlier work of Lee and Spong (2006) [14] to improve position tracking and consequently transparency in the face of disturbances and environmental contacts. The proposed controller employs a PID controller in each side to overcome some limitations of a PD controller and guarantee an improved performance. Moreover, by using Fourier transform and Parseval’s identity in the frequency domain, we demonstrate that this new PID controller preserves the passivity of the system. Simulation and semi-experimental results show that the PID controller tracking performance is superior to that of the PD controller tracking performance in slave/environmental contacts.

1. ISA Transactions 51 (2012) 74–80
Contents lists available at SciVerse ScienceDirect
ISA Transactions
journal homepage: www.elsevier.com/locate/isatrans
Bilateral control of master–slave manipulators with constant time delay
A. Forouzantabara,∗
, H.A. Talebib,1
, A.K. Sedighc,2
a
Department of Electrical Engineering, Science and Research Branch, Islamic Azad University, Tehran, Iran
b
Department of Electrical Engineering, AmirKabir University of Technology, Tehran, Iran
c
Department of Electrical Engineering, Khaje Nasir Toosi University of Technology, Tehran, Iran
a r t i c l e i n f o
Article history:
Received 6 November 2010
Received in revised form
8 July 2011
Accepted 27 July 2011
Available online 20 August 2011
Keywords:
Robotic
Time delay
Passivity
Teleoperation
Position coordination
Transparency
a b s t r a c t
This paper presents a novel teleoperation controller for a nonlinear master–slave robotic system with
constant time delay in communication channel. The proposed controller enables the teleoperation system
to compensate human and environmental disturbances, while achieving master and slave position
coordination in both free motion and contact situation. The current work basically extends the passivity
based architecture upon the earlier work of Lee and Spong (2006) [14] to improve position tracking
and consequently transparency in the face of disturbances and environmental contacts. The proposed
controller employs a PID controller in each side to overcome some limitations of a PD controller and
guarantee an improved performance. Moreover, by using Fourier transform and Parseval’s identity in the
frequency domain, we demonstrate that this new PID controller preserves the passivity of the system.
Simulation and semi-experimental results show that the PID controller tracking performance is superior
to that of the PD controller tracking performance in slave/environmental contacts.
© 2011 ISA. Published by Elsevier Ltd. All rights reserved.
1. Introduction
Over the past 3 decades, teleoperation technologies have
been gradually growing through the world. Teleoperation is
used in many applications such as space operation [1], handling
of toxic and harmful materials [2], robotic surgery [3] and
underwater exploration [4]. Teleoperation can be divided into
two main categories, namely, unilateral and bilateral. In unilateral
teleoperation, the contact force feedback is not transmitted to
the master. In bilateral teleoperation, the remote environment
provides some necessary information by many different forms,
including audio, visual displays, or tactile through the feedback
loop to the master side. However, the contact force feedback
(haptic feedback) can provide a better sense of telepresence and
as a consequence improve task performances [5].
There are many structures for the bilateral teleoperation sys-
tem. Two main structures are two-channel (2CH) architecture [6]
and four-channel (4CH) architecture [7,8]. In two-channel struc-
ture usually the master position is sent to the slave controller, and
the contact force of the slave robot with the environment is directly
transmitted to the master.
∗ Corresponding author. Tel.: +98 2144865100; fax: +98 9177013735.
E-mail addresses: a.forouzantabar@srbiau.ac.ir, Ahmad.foruzan@gmail.com
(A. Forouzantabar), alit@aut.ac.ir (H.A. Talebi), sedigh@kntu.ac.ir (A.K. Sedigh).
1 Tel.: +98 2164543340.
2 Tel.: +98 218846 2175x317.
In bilateral teleoperation, there are two main objectives that
ensure a close coupling between the human operator/master
robot and slave robot. The first goal is that the slave robot
tracks the position of the master robot and the other is that
the force, that occurs when the slave contacts with the remote
environment, accurately transferred to the master. When these
conditions are met, the bilateral teleoperation system is called
a transparent system. Lawrence [8] has shown that there is a
tradeoff between the stability and transparency, the improvement
of one will deteriorate the other. The delay existing in the network
teleoperation system can destabilize the closed-loop system and
degrade transparency.
Most previous studies on stability were based on the passivity
formalism, such as scattering theory [9] and wave variables [10].
The key point for these approaches is to passify the non-passive
communication medium with time delay. Although transparency
of these two approaches is poor, the stability is robust against the
communication delay and called the delay-dependent stability. A
comprehensive survey on the delay compensation methods can be
found in [11]. Chopra and Spong [12] proposed a new architecture
which builds upon the scattering theory by using additional
position control on both the master and slave sides. This new
architecture has an improved position tracking and comparable
force tracking abilities than the traditional teleoperator model
of [9,10].
In [13], Lee and Spong, introduced a PD-based controller scheme
for the teleoperation system that keeps position coordination
0019-0578/$ – see front matter © 2011 ISA. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.isatra.2011.07.005

2. A. Forouzantabar et al. / ISA Transactions 51 (2012) 74–80 75
and ensures the passivity of the closed-loop system. The main
drawback of this structure is that the backward and forward
communication delays must be exactly known and symmetric.
Therefore, they removed these aforementioned restrictions in their
recent works. They used the controller passivity concept, the
Lyapunov–Krasovskii technique, and Parseval’s identity, to passify
the combination of the delayed communication and control blocks
altogether since, the delays are finite constants and an upper bound
for the round-trip delay is known [14,15].
Nuno et al. [16] showed that it is possible to control a
bilateral teleoperation with a simple PD-like controller and achieve
stable behaviour under specific condition on control parameters.
According to the complexity of the communication network, the
backward and forward delays are not only time-varying but also
asymmetric. In [17,18], two different methods based on the PD
controller have been presented to address these problems. The
method in [18] uses a Lyapunov–Krasovskii functional to derive
the delay-dependent stability criteria, which is given in the linear-
matrix-inequality (LMI) form. Ryu et al. [19,20] proposed a passive
bilateral control scheme for the teleoperation system with time-
varying delay, which is composed of a Passivity Observer and
a Passivity Controller. This controller guarantees the passivity
of bilateral teleoperation under some condition, independent
of the amount and variation of time-delay in communication
channel. In [21], the authors also extended the previously proposed
controller in [14]. The main difference is the use of stable neural
network in each side to approximate the unknown nonlinear
functions in the robot dynamics and enhance the master–slave
tracking performance in the face of different initial conditions
and environmental contacts. The new neural network controller
preserved the passivity of the overall system.
In this paper, the passivity based architecture of [14] is ex-
tended to improve position tracking and consequently trans-
parency in the face of environmental contacts. In this regard, a PID
controller is employed in each side to overcome some limitations
of PD controllers such as disturbance rejection. The key feature of
the proposed PID controller is that it preserves the control passivity
of the teleoperation system. For this purpose, we will use Fourier
transform, Parseval’s identity, and the Schur complement to show
that the proposed PID controller with additional dissipation term
will preserve the passivity of the system under some mild condi-
tion, since the time delays are constant. The majority of the passiv-
ity demonstration is done in the frequency domain.
The rest of this paper is organized as follows. Section 2 describes
passive bilateral teleoperation structure with constant time delay
introduced by Lee and Spong [14]. In Section 3, we describe the new
control architecture based on the PID controller and demonstrate
its passivity. Furthermore, we used a Nicosia observer [23] to
estimate the human hand forces when the master does not have
force sensor. Section 4 shows the simulation and experimental
results. And finally Section 5 draws conclusions and gives some
suggestions for future works.
2. Passive bilateral teleoperation structure with constant time
delay
In [14], a novel control framework for bilateral teleoperation
of a pair of multi-DOF nonlinear robotic systems with constant
communication delays was proposed. The proposed bilateral
teleoperation framework is shown in Fig. 1.
A bilateral teleoperation system which is shown in Fig. 1
consists of five interacting subsystems: the human operator, the
master manipulator, the control and communication medium,
the slave manipulator and the environment. The human operator
commands via a master manipulator by applying a force F1(Fh)
to move it with position q1 and velocity ˙q1 which is transmitted
to the slave manipulator through the communication medium. A
local control (T2) on the slave side drives the slave position q2 and
velocity ˙q2 towards the master position and velocity. If the slave
contacts a remote environment and/or some external source, the
remote force F2(−Fe) is sent back from the slave side and received
at the master side as the force or control signal T1.
Assuming the absence of friction, gravitational forces and other
disturbances, the equation of motion for a master and slave
nonlinear robotic system is given as follows [14]
M1(q1)¨q1 + C1(q1, ˙q1)˙q1 = T1(t) + F1(t) (1)
M2(q2)¨q2 + C2(q2, ˙q2)˙q2 = T2(t) + F2(t) (2)
where qi(t) ∈ ℜn
are the vector of joint displacements, ˙qi(t) ∈
ℜn
are the vector of joint velocities, Ti(t) ∈ ℜn
are the control
signals, Fi(t) ∈ ℜn
represent the human/environmental force,
Mi(qi) ∈ ℜn×n
are symmetric and positive-definite inertia matrices
and Ci(qi, ˙qi) ∈ ℜn×n
are the Coriolis/centripetal matrices (i =
1, 2). The Lagrangian robot dynamics enjoys certain fundamental
properties [22].
• Property 1: The inertia matrix M(q) is symmetric, positive
definite and bounded so that
µ1I ≤ M(q) ≤ µ2I
where the bounds µi, i = 1, 2 are positive constants.
• Property 2: The Coriolis/centripetal matrices can always be
selected such that the matrix ˙M(q) − 2C(q, ˙q) is skew-
symmetric.
The control objectives are designing the controllers Ti(t), i =
1, 2 to achieve these two goals:
I. master–slave position coordination: if (F1(t), F2(t)) = 0,
qE (t) := q1(t) − q2(t) → 0, t → ∞ (3)
II. static force reflection: with (¨q1(t), ¨q2(t), ˙q1(t), ˙q2(t)) → 0
F1(t) → −F2(t) or Fh(t) → Fe(t). (4)
The closed-loop teleoperator (1)–(2) is said to satisfy the
energetic passivity condition if there exists a finite constant d ∈ ℜ
such that for t ≥ 0:
∫ t
0

FT
1 (θ)˙q1(θ) + FT
2 (θ)˙q2(θ)

dθ ≥ −d2
. (5)
The teleoperator controllers hold the controller passivity if there
exists a finite constant c ∈ ℜ such that for t ≥ 0:
∫ t
0

TT
1 (θ)˙q1(θ) + TT
2 (θ)˙q2(θ)

dθ ≤ c2
. (6)
This means that the energy generated by the master and
slave controllers is always bounded [15]. It is shown that for
teleoperation system (1)–(2), controller passivity (6) indicate
energetic passivity (5). This allows us to investigate the passivity
of the closed-loop teleoperator just by checking the controller
structure and with no concern about the nonlinear dynamics of the
master and slave robots [14].
The following PD-like controllers are proposed to guarantee the
master–slave coordination (3), bilateral force reflection (4), and
energetic passivity (5).
T1(t) = −Kv(˙q1(t) − ˙q2(t − τ2)) − (Kd + Pε)˙q1(t)
− Kp(q1(t) − q2(t − τ2)) (7)
T2(t) = −Kv(˙q2(t) − ˙q1(t − τ1)) − (Kd + Pε)˙q2(t)
− Kp(q2(t) − q1(t − τ1)) (8)
where τ1, τ2 ≥ 0 are the finite constant communication delays,
Kυ, Kp ∈ ℜn×n
are the symmetric and positive-definite PD gains,

3. 76 A. Forouzantabar et al. / ISA Transactions 51 (2012) 74–80
Fig. 1. The schematic of the closed-loop teleoperation system [14].
Pε ∈ ℜn×n
is an additional damping that guarantee master–slave
coordination (3), and Kd ∈ ℜn×n
is the dissipation to passify the
proportional term in the master and slave controllers. One of the
options for Kd is
Kd =
¯τrt
2
Kp (9)
where ¯τrt ≥ 0 is an upper-bound of the round-trip communication
delay which is the sum of the forward (τ1) and backward (τ2)
delays.
3. The proposed controller scheme
In order to eliminate the steady-state tracking error and
improve the disturbance rejection when the slave contact to the
environment, we extend a PD controller to a PID controller. The
integral term is added so that the system will have no steady-
state error in the presence of constant disturbances. The controller
design is mainly based on analysing the energy production of
the PID controller gains in frequency domain and passifies the
effect of these gains. In sequel, the passivity of the presented
control methodology is illustrated through rigorous mathematical
analysis.
Adding an integral term and additional dissipation term Kd2 to
(7), (8) control law, the PID control law can be introduced as
T′
1(t) = −Kv(˙q1(t) − ˙q2(t − τ2)) − Kp(q1(t) − q2(t − τ2))
− Ki
∫ t
0
(q1(t) − q2(t − τ2))dt − (Kd + Kd2
+ Pε)˙q1(t) (10)
T′
2(t) = −Kv(˙q2(t) − ˙q1(t − τ1)) − Kp(q2(t) − q1(t − τ1))
− Ki
∫ t
0
(q2(t) − q1(t − τ1))dt − (Kd
+ Kd2 + Pε)˙q2(t). (11)
The following theorem shows that the two local controllers
T′
1(t), T′
2(t) in (10), (11) guarantee energetic passivity (5) of the
closed-loop teleoperation system under the condition on Kd2 which
is given in (12). Note that Kp, Kv, Ki, Kd and Kd2 are positive-definite
diagonal matrices.
Theorem 1. Consider the nonlinear bilateral teleoperation system
given by (1), (2) with the controllers (10), (11). If condition (9) is hold
with the gain Kd2 satisfying
Kd2 ≥
2 cos2

ω(τ1+τ2)
4

ω2
Ki, ∀ω. (12)
Then the closed-loop teleoperation system is energetic passive.
Proof. The procedure of the passivity proof is similar to that
presented in [14] with the difference that our controller has extra
integral and dissipation terms. Using Fourier transform, Parseval’s
identity and under some mild condition on integral and dissipation
gains, it was shown that the presented controller preserve system
passivity.
Substituting (10), (11) in the definition of controller passivity
(6), we obtain
∫ t
0

T
/T
1 (θ)˙q1(θ) + T
/T
2 (θ)˙q2(θ)

dθ ≤ c2
∀t ≥ 0. (13)
Indicating the terms in integral (13) by Pt (t) and substituting from
(10), (11) yield
Pt (t) := Pd(t) + Pp(t) + Pi(t) − P(t) (14)
where Pd(t), Pp(t) and Pi(t) are the powers related to the delayed
D-action, delayed P-action (+ dissipation Kd) and delayed I-action
(+ dissipation Kd2), respectively and defined by
Pd(t) := −˙qT
1 (t)Kv ˙q1(t) + ˙qT
1 (t)Kv ˙q2(t − τ2)
− ˙qT
2 (t)Kv ˙q2(t) + ˙qT
2 (t)Kv ˙q1(t − τ1) (15)
PP (t) := −˙qT
1 (t)Kd ˙q1(t) − ˙qT
1 (t)Kp(q1(t) − q2(t − τ2))
− ˙qT
2 (t)Kd ˙q2(t) − ˙qT
2 (t)Kp(q2(t) − q1(t − τ1)) (16)
Pi(t) := −˙qT
1 (t)Kd2˙q1(t) − ˙qT
1 (t)Ki
∫ t
0
(q1(t) − q2(t − τ2))dt
− ˙qT
2 (t)Kd2 ˙q2(t)− ˙qT
2 (t)Ki
∫ t
0
(q2(t) − q1(t − τ1))dt (17)
and P(t) is the following positive-definite quadratic form:
P(t) :=

˙q1(t)
˙q2(t)
T [
Pε 0
0 Pε
] 
˙q1(t)
˙q2(t)

. (18)
It is shown in [14] that
∫ t
0
Pd(θ)dθ ≤ −Vv(t) + Vv(0) (19)
where Vv(t) ≥ 0 ∀t ≥ 0 is a Lyapunov–Krasovskii functional for
delayed systems defined by
Vv(t) :=
2−
i=1
1
2
∫ 0
−τi
˙qT
i (t + θ)Kv ˙qi(t + θ)dθ ≥ 0. (20)
Moreover, if condition (9) is satisfied, it is shown in [14] that
∫ t
0
Pp(θ)dθ ≤ −Vp(t) + Vp(0) (21)
where VP (t) ≥ 0 ∀t ≥ 0 is defined as
Vp(t) :=
1
2
qT
E (t)KpqE (t) (22)
with qE (t) = q1(t) − q2(t).

4. A. Forouzantabar et al. / ISA Transactions 51 (2012) 74–80 77
Considering the fact that qi(t), ˙qi(t) = 0 ∀t ∈ (−∞, 0] then,
the energy generation by Pi, the power in (17), can be written as
∫ t
0
Pi(λ)dλ
= −
∫ ∞
−∞
˙qT
1 (λ)Kd2˙q1(λ)dλ −
∫ ∞
−∞
˙qT
2 (λ)Kd2 ˙q2(λ)dλ
−
∫ ∞
−∞
˙qT
1 (λ)
[
Ki
∫ t
−∞
(q1(ξ) − q2(ξ − τ2))dξ
]
dλ
−
∫ ∞
−∞
˙qT
2 (λ)
[
Ki
∫ t
−∞
(q2(ξ) − q1(ξ − τ1))dξ
]
dλ. (23)
Now, denoting the Fourier transforms of qi(t), (i = 1, 2) by
Vi(jω) =
 ∞
−∞ qi(t)e−jωt
dt =
 t
0
qi(t)e−jωt
dt, and using Parseval’s
identity, we have
∫ ∞
−∞
˙qT
1 (λ)Kd2 ˙q1(λ)dλ
=
1
2π
∫ ∞
−∞
(−jω)V∗
1 (jω)Kd2(jω)V1(jω)dω
=
1
2π
∫ ∞
−∞
(ω2
)V∗
1 (jω)Kd2V1(jω)dω (24)
∫ ∞
−∞
˙qT
l (λ)
[
Ki
∫ t
−∞
(ql(ξ) − qk(ξ − τk))dξ
]
dλ
=
1
2π
∫ ∞
−∞
(−jω)V∗
l (jω)Ki
 1
jω
Vl(jω)
+ πVl(0)δ(ω) −
e−jωτk
jω
Vk(jω) + πVk(0)δ(ω)

dω
= −
1
2π
∫ ∞
−∞
V∗
l (jω)Ki[Vl(jω)
− e−jωτk
Vk(jω)]dω, (l, k) = {(1, 2), (2, 1)} (25)
where V∗
i is the complex conjugate transpose of a complex vector
Vi. The last equality is obtained by using the fact that ωδ(ω) = 0.
Then, substituting (24), (25) in (23) we have
∫ t
0
si(λ)dλ = −
1
2π
∫ ∞
−∞
ω2
V∗
1 (jω)Kd2V1(jω)dλ
−
1
2π
∫ ∞
−∞
ω2
V∗
2 (jω)Kd2V2(jω)dλ
−
1
2π
∫ ∞
−∞
V∗
1 (jω)Ki

e−jωτ2
V2(jω) − V1(jω)

dω
−
1
2π
∫ ∞
−∞
V∗
2 (jω)Ki

e−jωτ1
V1(jω) − V2(jω)

dω
= −
1
2π
∫ ∞
−∞
[
¯V1(jω)
¯V2(jω)
]T
H(jω)
[
¯V1(jω)
¯V2(jω)
]
dω (26)
where, H(jω) ∈ C2n×2n
is given by
H(jω) =



ω2
Kd2 − Ki
Ki
2
(ejωτ1
+ e−jωτ2
)
Ki
2
(e−jωτ1
+ ejωτ2
) ω2
Kd2 − Ki;


 (27)
for more details see [14]. Since H(jω) is Hermitian, then using the
Schur complement, H(jω) is a positive-semidefinite matrix if and
only if
ω2
Kd2 − Ki ≥ 0 → ω2
Kd2 ≥ Ki (28)
and
(ω2
Kd2 − Ki)
≥
(e−jωτ1 + ejωτ2 )
2
(ejωτ1 + e−jωτ2 )
2
Ki(ω2
Kd2 − Ki)−1
Ki
=
1 + cos ω(τ1 + τ2)
2
Ki(ω2
Kd2 − Ki)−1
Ki (29)
which is always true if
Kd2 ≥
2 cos2

ω(τ1+τ2)
4

ω2
Ki. (30)
Selecting the gain according to (30) guarantees that
 t
0
Pi(λ)dλ
is semi-negative. Since Vv(t) ≥ 0, Vp(t) ≥ 0 and P(t) ≥ 0, ∀t ≥ 0
therefore, by considering inequalities (19), (21) with (18), we have
∫ t
0
T′T
1 (θ)˙q1(θ) + T′T
2 (θ)˙q2(θ)
≤ −Vv(t) + Vv(0) − Vp(t) + Vp(0) −
∫ t
0
P(θ)dθ
≤ Vv(0) + Vp(0) =: c2
(31)
with c a finite constant. Thus the controller passivity is proved.
Finally, from the fact that, for teleoperation system (1)–(2), con-
troller passivity (6) results energetic passivity [14], energetic pas-
sivity (5) of the closed-loop teleoperation system is demonstrated.
It is also concluded that, if (¨q1(t), ¨q2(t), ˙q1(t), ˙q2(t)) → 0, then
from the master–slave robot dynamics (1), (2) and their controls
(10), (11), we get F1(t) → −F2(t) → −Kp(q1(t) − q2(t)) −
Ki
 t
0
(q1(t) − q2(t))dt where ˙qi(t − τi) → 0 and qi(t − τi) →
qi(t).
3.1. Hand force observer
In our master–slave system, master is a two-DOF joystick and
has uncoupled motions about the two axes due to its gimbal-based
design. In the experiment, the measurements of hand/master
forces F1(Fh) are required. Because our master do not have a force
sensor, we use a nonlinear state observer to estimate F1(Fh). Now,
choosing x1 = q1 and x2 = ˙q1, we write the master dynamics in
state-space as
˙x1 = x2
˙x2 = M−1
1 (q1)(−C1(x1, x2)x2 + T1(t) + F1(t)).
(32)
The Nicosia observer, which is used to estimate the hand forces
F1 or joint velocity ˙q1, described by [23,24]:
˙ˆx1 = ˆx2 + k2e
˙ˆx2 = M−1
1 (x1)(−C1(x1, ˙ˆx1)˙ˆx1 + T1(t) + K1e)
e = x1 − ˆx1
(33)
where ˆx1 and ˆx2 are the estimated joint position and velocity, e is
the output observation error, k2 is a positive scalar constant and
K1 is a symmetric positive-definite matrix. This nonlinear observer
uses joint position and the portion of the joint torque which comes
from the controller to estimate the external applied joint force. It
is shown in [23] that the observer is asymptotically stable and the
error dynamics is:
M1 ¨e + k2M1 ˙e + K1e = F1. (34)
In steady state, ¨e = ˙e = 0. In this result, at low frequency
operation or in steady-state the hand force is estimated as
proportional relationship to position estimation error as ¯F1 = K1e.

5. 78 A. Forouzantabar et al. / ISA Transactions 51 (2012) 74–80
Fig. 2. The master and the slave joint positions with the PID controller for 2-DOF
robots.
4. Simulation and experiment
4.1. Simulation
In order to evaluate the effectiveness of the proposed control
scheme in this paper, the controller has been applied to a pair of 2-
link planar RR robot arm. The robot dynamics was given in [25].
The lengths of serial links for both master and slave robots are
considered as a1 = a2 = 1.3 m, and inertia of the links was
taken as m1 = 0.8, m2 = 2.3 kg. To evaluate the system’s contact
behaviour, a virtual soft ball is installed in the slave environment
at x = 0.5. The ball is modelled like a spring–damper system
with the spring and damping gains as 100 N/m and 0.2 Ns/m. The
controller gains are chosen such that the close loop system behaves
as a critical damping system. In this regard, the gains Kp, Kv and Ki
are selected Kv = 40I2×2, Kp =
Kv
2
2
= 400I2×2, Ki = I2×2 and
the Pε is also set to zero. Note that, Ki has been chosen small so
that the third-order error dynamics is close to the second-order
error dynamics without this term (i.e., a dominant pole analysis
can be performed). The delays were selected as τ1 = τ2 =
1 s. Then, according to condition (9), (12) we choose (Kd, Kd2) =
(400I2×2, 5I2×2). Because the observer should have fast response,
the observer gains are taken as (K1, k2) = (450I2×2, 45). The
scenario is set up such that the operator moves the master robot
from the home position in x-direction to (x, y) = (1, 0) in the
Cartesian space and returns it to imitate a sinusoidal input. The
slave trying to follow these commands is steered towards an
obstacle. As a result, force information is generated and sent back
to the master.
The simulation result for the two joint positions and force
tracking are shown in Figs. 2 and 3. The force and position
tracking and consequently transparency are satisfactory. The
master position in the Cartesian space and the slave contact force is
also shown in Fig. 4. As shown in Figs. 3 and 4, when the slave robot
contact with the ball at x = 0.5, the contact force is reflected to the
master. Moreover, when the slave robot moves in free motion the
interaction force is zero.
4.2. Experiment
In the experiment, a two-DOF Logitech joystick with force
feedback is used as the master. The joystick has uncoupled motions
about the two axes due to its gimbal-based design. We use joystick
only in x-direction. The slave, which is constructed as a virtual
robot, is a 1_DOF robot with prismatic joint. The slave dynamics is
Fig. 3. The estimation of human force F1 and the environmental force F2 with the
PID controller for 2-DOF robots.
Fig. 4. The master position in the Cartesian space and the slave contact force.
assumed to be M2 = 0.7 kg, C2 = 2 Ns/m. The communication
delays and the environment model are the same as before. The
virtual slave robot and a ball as environment placed in front of it, is
shown in Fig. 5. The integration of the master–slave teleoperation
is accomplished through the MATLAB Simulink environment.
The scenario is set up such that the operator moves the joystick
away from the home position at x = 0 to the position x =
0.8 and returns it to imitate a step input. The slave trying to
follow these commands is steered towards an obstacle. As a
result, force information is generated and sent back to the master.
The controller gains are chosen such that the close loop system
behaves as a critical damping system. For this purpose, the control
parameters are chosen as Kv = 10, Kp =
Kv
2
2
= 25, Ki =
1 and the Pε is also set to zero. Then, According to (9), (12),
(Kd, Kd2) = (25, 5) are chosen. The observer gains are taken as
(K1, k2) = (250, 100). The experimental results for position and
force tracking are shown in Figs. 6 and 7.
To study the contribution of the proposed controller, we
implement the conventional controller (PD) (7), (8) with the
same control parameters. Figs. 8 and 9 show the results. The
position tracking performance is not good. However force tracking
is slightly better for the PID controller.
To sum up, it is clear that the proposed PID controller
makes a significant improvement in position tracking and rejects
the constant disturbance (environmental contact) as a result,
transparency is improved.

6. A. Forouzantabar et al. / ISA Transactions 51 (2012) 74–80 79
Fig. 5. Virtual slave robot with a ball as environment in free motion and contact situation.
Fig. 6. The master and the slave positions with the PID controller for 1-DOF robots.
Fig. 7. The estimated human force F1 and the environmental force F2 with the PID
controller for 1-DOF robots.
5. Conclusions
A novel PID control architecture for the bilateral teleoperation
system with time delay has been proposed in this paper that
ensures position coordination and static force reflection in the
presence of disturbances and environmental contacts. The new
architecture extends a PD controller to a PID controller that allows
the teleoperation system to reject the constant disturbance. We
have shown that the new PID controller preserve the energetic
passivity under the condition on additional dissipation term. Since
the master does not have a force sensor, a state observer is used
Fig. 8. The master and the slave positions with the PD controller for 1-DOF robots.
Fig. 9. The estimated human force F1 and the environmental force F2 with the PD
controller for 1-DOF robots.
to estimate the human force. The new controller provides better
transparency which is measured in terms of position and force
tracking ability of the bilateral system. Evaluating the performance
of this controller architecture with real slave robot remains for
future work.
References
[1] Imaida T, Yokokohji Y, Doi T, Oda M, Yoshikwa T. Groundspace bilateral
teleoperation of ETS-VII robot arm by direct bilateral coupling under 7-s time
delay condition. IEEE Trans Robot Autom 2004;20(3):499–511.

7. 80 A. Forouzantabar et al. / ISA Transactions 51 (2012) 74–80
[2] Wang W, Yuan K. Teleoperated manipulator for leak detection of sealed
radioactive sources. In: Proceedings of the IEEE international conference on
robotics and automation. vol. 2. 2004. p. 1682–7.
[3] Funda J, Taylor RH, Eldridge B, Gomory S, Gruben KG. Constrained Cartesian
motion control for teleoperated surgical robots. IEEE Trans Robot Autom 1996;
12(3):453–65.
[4] Funda J, Paul RP. A symbolic teleoperator interface for time-delayed
underwater robot manipulation. In: Ocean technologies and opportunities in
the Pacific for the 90’s. 1991. p. 1526–33.
[5] Chopra N, Spong MW, Lozano R. Adaptive coordination control of bilateral
teleoperators with time delay. In: IEEE conference on decision and control.
2004. p. 4540–7.
[6] Hannaford B. A design framework for teleoperators with kinesthetic feedback.
IEEE Trans Robot Autom 1989;5(4):426–34.
[7] Yokokohji Y, Yoshikawa T. Bilateral control of master–slave manipulators
for ideal kinesthetic coupling-formulation and experiment. IEEE Trans Robot
Autom 1994;10(5):605–19.
[8] Lawrence DA. Stability and transparency in bilateral teleoperation. IEEE Trans
Robot Autom 1993;9(5):624–37.
[9] Anderson RJ, Spong MW. Bilateral control of teleoperators with time delay.
IEEE Trans Automat Control 1989;AC-34(5):494–501.
[10] Niemeyer G, Slotine J. Stable adaptive teleoperation. IEEE J Oceanic Engineer-
ing 1991;16(1):152–62.
[11] Hokayem PF, Spong MW. Bilateral teleoperation: an historical survey.
Automatica 2006;42(12):2035–57.
[12] Chopra N, Spong MW, Ortega R, Barabanov NE. On position tracking in bilateral
teleoperation. In: American control conference. 2004. p. 5244–9.
[13] Lee DJ, Spong MW. Passive bilateral control of teleoperators under constant
time-delay. In: Proceedings of the IFAC world congress. 2005.
[14] Lee DJ, Spong MW. Passive bilateral teleoperation with constant time-delay.
IEEE Trans Rob 2006.
[15] Lee DJ, Spong MW. Passive bilateral teleoperation with constant time-delay.
In: IEEE int’l conference on robotics & automation. 2006.
[16] Nuno E, Ortega R, Barabanov N, Basanez L. A globally stable PD controller for
bilateral teleoperators. IEEE Trans Rob 2008;24(3):753–8.
[17] Hua C, Liu XP. Delay-dependent stability criteria of teleoperation systems with
asymmetric time-varying delays. IEEE Trans Rob 2010;26(5).
[18] Nuno E, Basanez L, Ortega R, Spong MW. Position tracking for non-
linear teleoperators with variable time delay. Int J Robot Res 2009;28(7):
895–910.
[19] Ryu JH, Preusche C. Stable bilateral control of teleoperators under time-varying
communication delay: time domain passivity approach. In: IEEE ICRA’07.
2007. p. 3508–13.
[20] Ryu JH, Artigas J, Preusche C. A passive bilateral control scheme for a
teleoperator with time-varying communication delay. Mechatronics 2010;
20(7):812–23.
[21] Forouzantabar A, Talebi H, Sedigh A. Adaptive neural network control of
bilateral teleoperation with constant time delay. Nonlinear Dynam 2011;1–12.
doi:10.1007/s11071-011-0057-8.
[22] Khalil HK. Nonlinear system. Prentice Hall; 2002.
[23] Nicosia S, Tomei P. Robot control by using only joint position measurements.
IEEE Trans Autom Control 1990;35(9):1058–61.
[24] Hacksel PJ, Salcudean SE. Estimation of environment forces and rigid-body
velocities using observers. In: Proceedings of the IEEE international conference
on robotics and automation. 1994. p. 931–6.
[25] Lewis FL, Jagannathan S, Yesildirek A. Neural network control of robot
manipulators and nonlinear systems. London: Taylor and Francis; 1999.