You are on page 1of 6

2005 IEEE/RSJ International Conference on Intelligent Robots and Systems

Multi-operator/Multi-robot Teleoperation: An Adaptive


Nonlinear Control Approach
Shahin Sirouspour and Peyman Setoodeh

Department of Electrical and Computer Engineering, McMaster University


Hamilton, Ontario, L8S 4K1, Canada
sirouspour@ece.mcmaster.ca, setoodp@mcmaster.ca

Abstract— Cooperative telerobotic systems consist of multiple arms to telemanipulate a surgical instrument). Despite ex-
pairs of master/slave robotic manipulators operating in a tensive amount of research in teleoperation, e.g. [2]–[8],
shared environment. This paper presents a multilateral adaptive and coordinated control of robots, e.g. [9], [10], multi-
nonlinear control architecture for cooperative teleoperation. The
proposed framework allows for transmission of position and master/multi-slave teleoperation has received little attention
force information between all master and slave robots rather in the past. This includes the work in [11] which proposes the
than merely between corresponding units. An adaptive non- use of heuristic methods such as predictive graphic displays
linear controller establishes kinematic correspondence among to address the time delay in the communication channels in
masters and slaves. The operators are presented with a virtual cooperative telerobotic environments. Approaches based on
intervening tool in order to collaboratively interact with the
environment. Models of operators, master and slave robots, tool, Petri Net models and event-based planning and control, have
and environment are incorporated in the design. The stability been recently developed in [12] for remote operation in multi-
of the system in the presence of parametric uncertainty in robot environments. An Internet-based distributed, multiple-
the dynamics is proven via Lyapunov analysis. Simulation and telerobot system that enables operators to use remote robots
experimental studies demonstrate that the proposed approach is in order to perform cooperative tasks has been introduced
highly effective in all phases of a teleoperation task, i.e. in free
motion, in contact with a flexible environment, and in contact in [13].
with a rigid environment. In cooperative teleoperation, multiple slave robots dynam-
Index Terms— Cooperative Teleoperation, Multi- ically interact, directly or through an intervening tool dynam-
master/Multi-slave Teleoperation, Robot Control, Nonlinear ics. Designs based on conventional single-master/single-slave
Control, Adaptive Control. architectures fail to address the performance and stability re-
quirements of cooperative teleoperation as they often neglect
I. I NTRODUCTION
these interactions. The issues of performance and stability for
Over the past two decades, applications of teleoperation such systems must be addressed under these constraints.
technology have been steadily growing in space and underwa- This paper follows upon the first author’s earlier work [14]
ter exploration, mining, toxic and nuclear material handling, in which a general architecture for cooperative teleoperation
the entertainment industry, and more recently in health care control was introduced. In that work, a µ-synthesis based
[1]. Telerobotic systems deliver the human intelligence and robust linear controller was developed assuming dynamically
manipulation skills combined with robot precision, repeata- linearized models for the operators, masters, slaves, and
bility and power to inaccessible or remote environments. This environment. The controller utilizes all possible communi-
is achieved through coordinated control of a master robotic cation channels between the master and slave manipulators.
arm, locally used by the operator, and a slave manipulator Alternatively, this paper proposes an adaptive nonlinear con-
which mimics the operator’s actions in the task environment. troller to accommodate nonlinearity and parametric uncer-
Cooperative robotic manipulation has advantages over tainty in the system dynamics. The control laws are inspired
single-robot manipulation, such as increased dexterity, im- by those given in [7] for conventional single-master/single-
proved handling capability, increased loading capacity, and slave teleoperation. The major contribution of this paper is
enhanced robustness due to redundancy. Cooperative tele- extending such control approach to cooperative teleoperation
operation combines advantages of human collaboration and applications where multiple masters and slaves are involved.
robotic manipulation. It can be employed in traditional tele- To the best knowledge of the authors, this is the first time that
operation applications as well as emerging areas such as a provably stable adaptive nonlinear controller for cooperative
robot-assisted surgery (e.g. a surgeon may use two robotic teleoperation is reported.
∗ This work was supported by National Sciences and Engineering Re-
The proposed multilateral teleoperation controller estab-
search Council of Canada, Canada Foundation for Innovation, and Ontario lishes position-position kinematic correspondence between
Innovation Trust. the masters and slaves. It also synthesizes an adjustable

0-7803-8912-3/05/$20.00 ©2005 IEEE. 2506


intervening tool impedance between the operators and the
environment. The stability of the closed-loop system in free
motion, and in contact with flexible and rigid environments
is demonstrated via the Lyapunov analysis. The novelty of
the proposed cooperative controller is due to: (i) explicitly
addressing the issues of performance and stability in the
presence of dynamic interaction between the slaves as well as
in the presence of unknown operators and the environment
dynamics. This is in contrast with the few reported work
in cooperative teleoperation which either do not consider
such dynamic interactions or only study the problem of free
motion coordination. (ii) its data communication architecture
which allows for all possible information routs.
The rest of this paper is organized as follows. Dynamics of Fig. 1. Schematic of a cooperative teleoperation system.
a cooperative teleoperation system is introduced in Section II. i’th master and operator can be integrated as follows
The adaptive nonlinear controllers for master and slave robots
are briefly discussed in Section III. The coordinating teleop- Mmh
i
Φ̈i (xim )+Cmh i
(Φi (xim ), Φ̇i (xim ))Φ̇i (xim )+Gimh (xim )
eration control laws are presented in Section IV. Experimental  
results for a single-axis two-master/two-slave design example = Ymhi
Φ̈i (xim ), Φ̇i (xim ), Φ̇i (xim ), Φi (xim ) θmh
i

are given in Section V. The paper is concluded in Section VI. = f¯m


i
+ f¯h∗i (1)

II. C OOPERATIVE T ELEOPERATION with xim ∈ R6 being the generalized position of a frame
attached to the i’th master’s handle w.r.t. its base frame; f¯m i

Figure 1 displays a conceptual cooperative teleoperation is the transformed control action and f¯h∗i is the operator’s
system in which m operators collaborate in performing a task intentional force, which is assumed to be bounded, i.e.
through controlling m slave robots and a tool. The operators |f¯h∗ij | < fmax
ij
; Φi (xim ) is the generalized master’s position
can move the tool in free motion, interact with a flexible transformed to the virtual tool frame. In (1), Ymh i
θmh
i
is the
environment, and make contact with a rigid environment. linear-in-parameter representation of the dynamics with θmh i

Cooperative teleoperation could also include applications in being the vector of unknown parameters of the operator and
which a single operator employs two hand controllers and possibly the master arm [15].
two robotic arms to complete a task. It is assumed that the slave robots are rigidly holding
In [14], it has been proposed to establish data links the tool and have formed a closed kinematic chain. This
between all master and slave units. Within such framework, assumption imposes the following constraint on the positions
each master and slave robot can receive position and force of the slaves and the tool.
information from all other master and slave units. The extra xt = Ψ1 (x1s ) = Ψ2 (x2s ) = · · · = Ψm (xm
s ) (2)
communication links can be utilized to facilitate coordination
among operators. For example, the controller can assist the where xt is the generalized position of a frame attached to
operators in simultaneous grasping of an object through the tool at its potential contact point with the environment, xis
imposing virtual constraints on positions of the slaves and is the generalized position of a frame attached to the contact
masters such that the slaves always keep the same distance point of the i’th slave and tool represented in the slaves’ base
from the object. Using this architecture, it is also possible frame and Ψ’s are appropriate coordinate transformations.
to impose a virtual intervening tool dynamics between the The tool-environment contacts are considered to be rigid
operators and the environment that they are working on. along some a priori known directions and flexible along the
Another advantage of the proposed architecture over tradi- rest. To simplify the analysis and without loss of generality,
tional approaches is in applications where communication it is further assumed the slaves’ base frame has been selected
delays are different throughout various channels. The local such that rigid and flexible coordinates are decoupled in
information links can be utilized to reduce instability due to generalized tool position vector xt . Therefore, xt can be
links with large delays. written as xt = ∆r xre + (I − ∆r ) xt where xre is a vector
The i’th master’s rigid body dynamics are governed by of constant contact positions along the rigid directions and
a second-order nonlinear differential equation and the op- ∆r is a 6 × 6 diagonal matrix that its i’th diagonal element
erator’s arm dynamics are approximated by a second-order is one when there is rigid contact in the i’th direction and
linear time-invariant differential equation. When the operator otherwise is zero. Hence, ∆r ẋt = 0 and ∆r ẍt = 0, and the
holds the master handle, i.e. xim = xih , the dynamics of the environment reaction force fe can be written in terms of its

2507
flexible and rigid components as follows action is Fs ∈ Rmn while xt ∈ Rn , Fsint is an internal force
component that does not affect the motion of the tool, i.e.
fe = ∆r fe + ∆f (Me ẍt + Be ẋt + Ke xt ) (3)
Jt Fsint = 0. This can help maintain the contact between the
where fe is a generalized force exerted on the environment slave manipulators and the tool if the end-effectors are not
by the tool at its contact point and ∆r fe is a vector of en- equipped with grippers.
vironment reaction forces along the rigid coordinates. Again The slave/environment parameter adaptation is driven by
∆f is a 6 × 6 diagonal matrix that its i’th diagonal element is ˙
θ̂ste = Γs Yste
T
(v̇tr , vtr , xt , xt ) ρt = Γs yθ (6)
one when there is flexible contact in the i’th direction and is
zeros otherwise; Me , Be , Ke ≥ 0 are constant matrices that where Γs is a diagonal matrix with elements
model the dynamics of the flexible part of the environment.  i
Since the rows and columns associated with rigid coordinates 
0 θ̂si ≤ θsmin , yθi ≤ 0
are zero, these matrices are positive semi-definite. γ ii = 0 θ̂si ≥ θsmax i , yθi ≥ 0 (7)

 i
Similarly, the dynamics of slave robots are governed by γs otherwise
second-order nonlinear differential equations and tool dynam-
ics are governed by a second-order linear differential equa- γ i ’s are positive and θsmin and θsmax are known lower and
tion. Dynamics of the slave robots, tool and the environment upper bounds on the parameters.
(3) can be combined to obtain The Control command for the i’th master is computed
using
 
Mste ẍt + Cste (xt , ẋt )ẋt + Gste =
fm
i
= Ymh
i
v̇m , vm , Φ̇i (xim ), Φi (xim ) θ̂mh
ri ri i
+ Km ρm
i i

m
T
Yste (ẍt , ẋt , ẋt , xt ) θste = Jti fsi − ∆r fe (4) +fmax
i
sgn(ρim )
i=1
vm
ri
= vm
di
+ Kp−1 Aim f˜hi , ρim = vm
ri
− Φ̇i (xim ) (8)
where the vector θste includes unknown parameters of the
flexible environment and possibly the slave robots and the Km i
, Λ, Kp , Am > 0 are diagonal and f˜hi is a filtered hand
tool; fsi is the equivalent force. The i’th master/operator parameter adaptations are
  control action in the slaves’ base
−1 given by
∂Ψi
frame and Jti = ∂xis .  
˙i
θ̂mh = Γim YmhT
v̇m , vm , Φ̇i (xim ), Φi (xim ) ρim = Γim zθi (9)
ri ri
III. A DAPTIVE N ONLINEAR C ONTROL FOR M ASTERS
AND S LAVES where Γim is a diagonal matrix with elements similar to those
The first step in deriving the cooperative teleoperation of Γs .
controllers is to design adaptive nonlinear control laws for B. Stability Analysis
master and slave robots. These controllers are similar to
those in [7] with two exceptions: First, there are multiple The following closed-loop error dynamics are obtained
slave manipulators and a tool here which form a closed for the slaves/tool/evironment and the i’th master/operator
kinematic chain and therefore their dynamics are coupled. subsystems by substituting control laws in (5) and (8) into
In [7] only one slave was considered. Second, Reference (4) and (1), respectively
   
[7] divides the contact space into two rigid and flexible Mste + ∆r C −1 A−1
s ρ̇t + Cste + ∆r A−1
s + K s ρt
subspaces. The modelling in this paper is slightly different
−Yste θ̃ste = 0 (10)
as it permits several flexible and rigid contact coordinates. In
other words, all flexible or rigid contacts do not necessarily
Mmh
i
(Φi (xim ))ρ̇im + Cmh
i
(Φi (xim ), Φ̇i (xim ))ρim +
occur simultaneously.
Km ρm + f¯h∗i + fmax
i i i
sgn(ρim ) − Ymhi
θ̃m
i
=0 (11)
A. Control Laws
with θ̃ste = θste − θ̂ste and θ̃m
i
= θmi
− θ̂m
i
. Candidate
The control laws for the slave robots are given by
Lyapunov functions for the error dynamics in (10) and (11)
 T
Fs = Jt† us + Fsint , Fs = fs1 · · · fsm are
us = Yste (v̇tr , vtr , ẋt , xt ) θ̂ste + Ks ρt + 1   1
 −1 d  Vs = ρTt Mste + ∆r C −1 A−1 ρt + θ̃sT Γs −1 θ̃s (12)
2 s
2
A−1
s ∆r C v̇t + vtd
v r = v d − As f˜e , ρt = v r − vt
t t t (5) Vmi =
1 i T i i
ρ Mmh ρm + θ˜m
1 i T i −1 i
Γm θ̃m (13)
2 m 2
and Ks , C, As > 0 are diagonal matrices, vtd is tool velocity
It is not difficult to show that
command, f˜e is a filtered environment reaction force that  
will be introduced in the next section. The slaves’ control V̇s ≤ −ρt ∆r A−1s + K s ρt (14)

2508
T
V̇mi ≤ −ρim Km ρm
i i
(15) which yields the following results

The results in (12)-(15) yield Φj (xjm ) − Φi (xim ) , Φ̇j (xjm ) − Φ̇i (xim ) ∈ L2 L∞ (25)

ρim = vm
di
− Φ̇(xim ) + Aim f˜hi ∈ L2 L∞ , i = 1, · · · , m and therefore the position tracking error between the i’th and

ρt = vtd − vt − As f˜e ∈ L2 L∞ (16) j’th master devices is stable. Furthermore,
Kp m k 1 m 
The switching terms in the masters’ control law (8) elim- ρt − Σk=1 ρm = Σk=1 Kp Φ̃˙ k (xkm ) − Ψ̃ ˙ k (xk ) +
inate the error due to unknown bounded external force fh∗ i .  m 2m 
s

However, high frequency switching activities are undesirable Kp m Kp m
Σ Φ̇ (xm ) − vt + Λ
k k
Σ Φ (xm ) − xt
k k
because they can excite unmodelled dynamics and cause m k=1 m k=1
instability. They can also negatively impact the life of the 1 
+ ΛΣm K p Φ̃ (x
k k
) − Ψ̃ (x
k k
) (26)
actuators. In practice, this can be resolved by replacing the 2m k=1 m s

sgn() with a smooth function or alternatively, by including K


Again given that ρt − mp Σm k=1 ρm ∈ L2
k
L∞ and following
fh∗ in θm and estimating it in real-time. The latter approach
similar steps to (21)-(23), it can be shown that
is adopted in this paper.
Kp m Kp m
IV. T ELEOPERATION C OORDINATING C ONTROLLERS Σ Φk (xkm ) − xt , Σ Φ̇k (xkm ) − vt
m k=1 m k=1 
Local adaptive nonlinear controllers in (5) and (8) guaran-
∈ L2 L∞ (27)
tee the stability of ρim and ρt . The operation of masters and
slaves are coordinated through the design of desired velocity Therefore, the stability of tracking errors between scaled
commands vm di
in (8) and vtd in (5) as follows: masters’ and slaves’ positions is guaranteed.
Λ 
Next, the force tracking behavior of the proposed cooper-
vmdi
= Σm Φ̃ (x
k k
) − Φ (x
i i
) − Kp−1 As f˜e
2m − 2 k=1,=i m m
ative teleoperation controller is studied.
Kp−1 
+ −1 m
k=1 Ψ̃ (xs ) − Kp Φ (xm ) +Kp Σk=1,=i Am fh
ΛΣm k k i i k ˜k Kp m k Kp m ˙ k k k 
2m ρt + Σk=1 ρm = Σk=1 Φ̃ (xm )vm − Φ̇k (xkm )
−1  m m

+
1
Σm ˙ k (xk ) + Kp Σm Ψ̃
Φ̃ ˙ k (xk ) (17) 1 m ˙k k Kp 
2m − 2 k=1,=i m
2m k=1 s + Σk=1 Ψ̃ (xs ) − vt + ΛΣm Φ̃ (x
k k
) − Φ (x
k k
) +
m m k=1 m m
   
Kp m ˙ k k 1 m 1 m 1 m ˙ k (xk ) − Ψ̃˙ k (xk )

vtd = Σk=1 Φ̃ (xm )+Kp Λ Σk=1 Φ̃k (xkm ) − Kp−1 xt Λ Σk=1 Ψ̃k (xks ) − xt + Σk=1 Kp Φ̃
m m m 2m m s

k ˜k 1 
+ Σm k ˜k ˜
k=1 Am fh (18) + ΛΣm K Φ̃ (x ) − Ψ̃ (xs +2Σk=1 Am fh −2As fe
)
k k k k m
p
2m k=1 m

In the above formulation (28)


˙ It should be noted that
Φ̃˙ k (xkm ) + C Φ̃k (xkm ) = CΦk (xkm ) , f˜hk + C f˜hk = C f¯hk ˙ k (xk )
˙ k (xk ) + C Ψ̃k (xk ) = CΨk (xk ) , ˙ Φ̃k (xk ) − Φk (xk ) = −C −1 Φ̃
f˜e + C f˜e = Cfe(19)
m m m
Ψ̃ s s s ˙ k (xk )
−1
Ψ̃
k
(xks )− xt = −C Ψ̃ s
Using (5),(8),(17), and (18), it can be shown ˙Φ̃k (xk ) − Φ̇k (xk ) = −C −1 Φ̃ ¨ k (xk )
m m m
  ˙ k (xk ) − v = −C −1 Ψ̃
¨ k (xk )
ρim − ρjm = Λ Φj (xjm ) − Φi (xim ) + Φ̇j (xjm ) − Φ̇i (xim )+ Ψ̃ s t s (29)
1  
˙ i (xi ) + Λ Φ̃j (xj ) − Φ̃(xi ) ]
[Φ̃˙ j (xjm ) − Φ̃ and also from (26),
2m − 2 m m m

(20) 1
Σ m
K Φ̃ ˙ k (xk ) ∈ L L
˙ k (xk ) − Ψ̃
Equation (20) can be written as 2m k=1 p m s 2 ∞

1 1
2m ΛΣk=1 Kp Φ̃ (xm ) − Ψ̃ (xs ) ∈ L2 L∞
m k k k k
ρim − ρjm = Ω̃ij + Ωij (21) (30)
2m − 2
Using (28)-(30) results in
 
Ω = Φ̇
ij j
(xjm ) − Φ̇ i
(xim ) + Λ Φj (xjm ) − Φi (xim ) (22)
˜k ˜ Kp C −1 m ¨ k k
k=1 Am fh − 2As fe −
2Σm Σk=1 Φ̃ (xm )−
˙ ij + C Ω̃ij = CΩij m
Ω̃ (23) C −1 m ¨ k k Kp C −1 Λ m ˙ k k
Σk=1 Ψ̃ (xs ) − Σk=1 Φ̃ (xm )−
Using (21), (23) and the fact that ρim − ρjm ∈ L2 L∞ , m m
C −1 Λ m ˙ k k 
 Σk=1 Ψ̃ (xs ) ∈ L2 L∞ (31)
Ωij ∈ L2 L∞ (24) m

2509
Note that
As f˜e = As ∆r f˜e + As ∆f f˜e (32)
and from (5) and (18)

As ∆r f˜e = ∆r vtd − ∆r ρt = ∆r Σm ˜k
k=1 Am fh − ∆r ρt − Λxt
Kp ˙k Kp Λ
+ ∆r Σm k=1 Φ̃ (xm ) + ∆r Σmk=1 Φ̃ (xm ) (33)
k k
m m
and using (3)
 
As ∆f f˜e = As Me x̃¨t + Be x̃˙ t + Ke x̃t (34)

¨ k (xk ) − B k Φ̃˙ k (xk ) − K k Φ̃k (xk ) (35)


f˜hk = f˜h∗k − Mhk Φ̃ m h m h m
Fig. 2. The cooperative teleoperation experimental setup.
f˜˙h∗k + C f˜h∗k = C f¯h∗k (36) twin pantograph haptic interfaces, each capable of producing
The following conclusions are drawn by substituting (32)- three active degrees of motion. The motion of each device is
(35) into (31), restricted to a single axis via the application of a workspace
proportional-derivative controller. Disturbance observers have
Φ̃k (xkm ) , Φ̃˙ k (xkm ) , Φk (xkm ) , Φ̇k (xkm ) ∈ L∞ (37) been designed and implemented to estimate the operators’
˙
Ψ̃ (x ) , Ψ̃ (x ) , Ψ (x ) , Ψ̇ (x ) ∈ L
k k k k k k k k
(38) hand forces rather using actual force sensors.
s s s s ∞
The control algorithm has been implemented using Matlab
This completes the proof of stability for the closed-loop Realtime Workshop Toolbox and Tornado 2.2/VxWorks 5.4
system. Using a rigorous analysis similar to that in [7], it RTOS by WindRiver. The control rate is set to 1024Hz.
is possible to show that ρim ’s and ρt converge to zero. The Two sets of experiments were conducted, one with flexible
convergence of position tracking errors to zero can then be contact and another with rigid contact. In the experiments,
concluded. two operators cooperatively moved the tool in free motion
Now by letting Am = Kf Af and assuming that the and made contact with the environment several times. Ex-
frequency of operation is below the bandwidth of the first- periments were repeated with different operators and similar
order filter C/2π, (34) can be approximated by behavior was observed.
¯k −1 A. Flexible Environment
k=1 fh − Kf fe
Σm
 Experimental results for flexible contact are shown in
≈ Kf−1 Kp A−1 C −1 Φ̈(xm ) + ΛΦ̇(xm ) (39) Figure 3. The system demonstrates accurate position tracking
both in free motion and in contact with the environment. In
This can be interpreted as a virtual intervening tool dy-
contact situation, the force exerted on the environment by
namics with a mass of Kf−1 Kp A−1 C −1 and damping of
the tool, tracks the operators’ net force after the velocity of
Kf−1 Kp A−1 C −1 Λ acting between the scaled environment
the tool approaches zero. This is expected behavior due to
force Kf−1 fe and the operators’ net force Σm ¯k
k=1 fh . The mass the intervening tool dynamics. Also, the adaptive controller
and stiffness of virtual tool are adjustable by the control is able to estimate the environment stiffness effectively as
parameters A, C, Kp , Kf , and Λ. shown in Figure 3(c). The small non-zero stiffness estimates
V. E XPERIMENTAL R ESULTS in free motion are most likely due to environment force
observation errors.
This section presents the results of experimental studies
conducted for evaluation of the proposed cooperative teleop- B. Rigid Environment
eration scheme on a single-axis linear two-master/two-slave Experimental results for the rigid contact are presented
system. The experimental setup is displayed in Figure 2. in Figure 4. The rigid contact behavior is stable and well
The two side carts on a linear track are motorized and are damped. The position tracking is accurate both in free motion
used as the slaves in the experiments. The middle cart can and in contact phases. The system also demonstrates excellent
freely move on the track and is employed as the tool. Its force tracking between the operators’ net force and the
motion is blocked in one direction by a rigid stop inserted environment force when the tool is pushed against the rigid
in the middle of the track under the tool. The position of the environment. Keeping the tool in contact with the rigid
slaves are measured by rotary optical encoders. Two force environment, oscillatory forces were intentionally exerted
sensors measure the interaction force between the slaves by operators to evaluate the force tracking ability of the
and the tool. The master devices are two identical planar controller.

2510
applications, this paper studied the problem of coordinated
control of multi-master/multi-slave teleoperation. The few
previous reports in cooperative teleoperation either ignore
the dynamic interaction of the slaves through the tool or
limit their scope to free motion coordination. This paper
incorporated the dynamics of master and slave manipulators,
operators, tool, and environment into an adaptive nonlinear
control approach. The stability of the closed-loop system in
the presence of parametric uncertainty and in contact with
flexible and rigid environments was proven through anal-
ysis. The controller establishes a desired position-position
correspondence between masters and slaves. It also pro-
vides the operators with a controllable virtual intervening
tool impedance. Simulation and experimental results demon-
strated that the proposed approach is highly effective in
coordinating the operation in a two-master/two-slave single-
axis teleoperation setting.
R EFERENCES
[1] C. Melchiorri and A. Eusebi, “Telemanipulation: system aspects and
control issues ,” in Proc. Model. Cont. Mechan. Robot., pp. 149–183,
Fig. 3. Experimental results for flexible environment: (a) position tracking 1996.
(b) force tracking (c) environment’s stiffness estimate. [2] B. Hannaford, “A design framework for teleoperators with kinesthetic
feedback,” IEEE Trans. Robot. Automat., vol. 5, no. 4, pp. 426–434,
1989.
[3] R. Anderson and M. Spong, “Bilateral control of teleoperators with
time delay,” IEEE Trans. Automat. Contr., vol. 34, no. 5, pp. 494–501,
1989.
[4] D. Lawrence, “Stability and transparency in bilateral teleoperation,”
IEEE Trans. Robot. Automat., vol. 9, no. 5, pp. 624–637, 1993.
[5] Y. Yokokohji and T. Yoshikawa, “Bilateral control of master-slave ma-
nipulators for ideal kinesthetic coupling-formulation and experiment,”
IEEE Trans. Robot. Automat., vol. 10, no. 5, pp. 605–620, 1994.
[6] G. Niemeyer and J.-J. Slotine, “Towards force-reflecting teleoperation
over Internet,” in IEEE Int. Conf. Robot. Auto., pp. 1909–1915, 1998.
[7] W.-H. Zhu and S. Salcudean, “Stability guaranteed teleoperation: an
adaptive motion/force control approach,” IEEE Trans. Automat. Contr.,
vol. 45, no. 11, pp. 1951–1969, 2000.
[8] R. Jee-Hwan, K. Dong-Soo, and B. Hannaford, “Stable teleoperation
with time-domain passivity control,” IEEE Trans. Robot. Automat.,
vol. 20, no. 2, pp. 365–373, 2004.
Fig. 4. Experimental results for rigid environment: (a) position tracking [9] J. Wang, S. Dodds, and W. Bailey, “Coordinated control of multiple
(b) force tracking. robotic manipulators handling a common object-theory and experi-
ments,” IEEE Proc. Contr. Theory Appl., vol. 144, pp. 73–84, January
Experiments have also been conducted using two decou- 1997.
pled four-channel teleoperation systems for performing the [10] R. Bonitz and T. Hsia, “Internal force-based impedance control for
same task. The system demonstrated poor contact stability. cooperating manipulators ,” IEEE Trans. Automat. Contr., vol. 12,
pp. 78–89, February 1996.
The operators were mostly unsuccessful in grasping the tool [11] N. Chong, T. Kotoku, K. Ohba, K. Komoriya, K. Tanie, J. Oaki,
because of contact instability. Even in cases in which the H. Hashimoto, F. Ozaki, K. Maeda, and N. Matsuhira, “A collaborative
force gains were reduced to improve stability, the contact multi-site teleoperation over an ISDN,” Mechatronics, vol. 13, pp. 957–
979, 2003.
between the tool and slaves were frequently lost due to poor [12] I. Elhajj, A. Goradia, N. Xi, C. M. Kit, Y.-H. Liu, and T. Fukuda,
coordination between the operators. In summary, the coop- “Design and analysis of internet-based tele-coordinated multi-robot
erative controller provides a superior transparent response systems,” Autonomous Robots, vol. 15, pp. 237–254, 2003.
[13] X.-G. Wang, M. Moallem, and R. Patel, “An internet-based distributed
in contact with the rigid environment and greatly facilitate multiple-telerobot system,” IEEE Trans. Syst., Man, Cybern. A, vol. 33,
the coordination between the operators when compared with no. 5, pp. 627–633, 2003.
decoupled single-slave/single-master controllers. [14] S. Sirouspour, “Robust control design for cooperative teleoperation,”
in Proc. IEEE Int. Conf. Robot. Auto., pp. 1145–1150, 2005.
[15] L. Sciavicco and B. Siciliano, Modeling and Control of Robot Manip-
VI. C ONCLUSIONS ulators. Springer, 2000.
While the vast majority of relevant reports in the literature
are concerned with single-master/single-slave teleoperation

2511

You might also like