Sunteți pe pagina 1din 13

12

Contact Control Concepts in


Manipulation Robotics-An Overview
Miomir VukobratoviC, Senior Member, IEEE, and Atanasko Tuneski

the same control strategies and techniques developed for the


control of robot motion in free space. In reality, however,
none of these conditions can be fulfilled. Hence, contact tasks
are characterized by dynamic interaction between robot and
environment, which often cannot be predicted accurately. In
many cases, the magnitude of mechanical work exchanged
between robot and environment during contact may vary
drastically and cause significant performance alteration for the
robotic control system.
Therefore, for successful completion of contact tasks, either
the interaction forces have to be monitored and controlled,
or control concepts ensuring compliant interaction with the
environment must be applied. Compliance can be considered
I. INTRODUCTION
as a measure of the ability of a manipulator to react on
HE nature of interaction between a robot and its en- interaction forces. The term refers to a variety of different
vironment can be categorized in two classes. The first control methods in which the end-effector motion is modified
one concerns the noncontact, e.g., unconstrained, motion in a by contact forces.
free work space, without any relevant environmental influence
The increasing demand for advanced robot applications has
exerted on the robot. In noncontact tasks, the robots own brought about an enormous growth of interest in the develdynamics have a crucial influence upon its performance. A opment of different concepts and schemes for the control of
limited number of the most frequently performed simple compliant motion. Strong support for the study of this problem
robotic tasks in practice, such as pick-and-place, spray paint- is derived from the new complex and somewhat unspecified
ing, gluing, or welding, belong to this group. In contrast to robotic tasks such as assembly, machining, or remote handling
these tasks, many complex advanced robotic applications such in an unstructured environment, where automation currently
as assembly and machining require the manipulator to be represents one of the most challenging topics in robotics.
mechanically coupled to other objects. These tasks are referred
The type of contact tasks may vary substantially for specific
to as essential contact tasks because they include phases where requirements, but in all cases, the robot has to perform three
the robots end-effector must come into contact with objects kinds of motion: so-called gross motion, related to robot
in its environment, produce certain forces upon them, and movement in free space; compliant or fine motion, related to
move along their surfaces. Inherently, each manipulation task robot movement constrained by an environment; and interface
requires contact with the object being manipulated. Many or approach motion, representing all transitions between gross
other tasks without prime emphasis on physical contact with and compliant motion.
the environment, such as inspection, might be considered
In this paper, we consider the current state of the art in
as potential contact tasks too. Motion through unstructured, constrained motion control.
insufficiently known workspace might also fall into this class.
The plan of this paper is as follows. In Section 11, the
The terms constrained or compliant motion are usually referred mathematical model of a robot, the model of actuators driving
to contact tasks.
the robots joints, and the environment model are presented.
Common to all the above-mentioned contact tasks is the In Section 111, the classification of different concepts for conpresence of motion constraints due to environmental objects. strained motion control is presented using several criteria. Two
Supposing all parameters of the environment and robot to be basic kinds of compliance will be distinguished and analyzed
known, and robot positioning to be ideally precise, it might separately in Sections IV and V. Section VI covers some
be possible to accomplish the majority of these tasks using recent stability issues. In Section VII, some experimental tests
of hybrid and impedance control are presented, while in the
Manuscript received April 30, 1993; revised September 27, 1993.
M. VukobratoviC is with the Robotics Laboratory, Mihailo Pupin Institute, concluding Section VlII, some of possible future investigation
1loo0 Belgrade, Yugoslavia.
subjects are indicated, and a unified approach to control law
A. Tuneski is with the Mechanical Engineering Faculty, St. Kin1 and
synthesis for robotic manipulators in contact with dynamic
Metodij University, 91000 Skopje, Macedonia.
IEEE Log Number 9214343.
environments is briefly reported.

Abstract-This paper presents the state of the art in the control


of robotic manipulators in constrained motion tasks. Contact
control concepts are classified using different criteria, and their
main characteristics are analyzed. For each of the presented
contact control concepts, the essential characteristics are stated.
The paper covers some early ideas and their later improvements,
as well as recent trends in this field. The advantages and drawbacks of the various control schemes are outlined, and they are
compared from the standpoint of their implementation issues. In
the paper, all characteristic results in the stability analysis of
robotic manipulators in the constrained motion tasks are briefly
reported. In the Conclusion, a new approach to the correct
solution of contact tasks control is mentioned as well.

02784046/94$04.00 0 1994 IEEE

VUKOBRATOVIC AND TUNESKI: CONTACT CONTROL CONCEPTS IN


11. MATHEMATICAL
MODELOF A ROBOT
IN

CONSTRAINED
MOTIONTASKS

Here, we shall briefly outline the main relations describing


the robots behavior in the contact tasks.
The mathematical model of the robot in the constrained
motion tasks consists of the model of the robots mechanism,
the model of the actuators driving its joints, and the contact
force model (environment model) [l], [ 2 ] .
A dynamic model of the robot mechanism with n DOFs
can be written in the form
Hq(q) . i

+ hq(q, d + g q k ) = pq - J T ( d. F

13

ROBOTICS

(1)

where s is the Laplace operator, K E is the (nx n ) semi-definite


environment stiffness matrix, DE is the (n x n ) semi-definite
environment damping matrix, and M E is the ( n x n ) positive
definite inertia matrix.
Consequently, the whole mathematical model of the robot
in the constrained motion tasks is given by the model of the
robots mechanism (l), the model of the actuators (2), and the
environment model (3), (4), or (5).
111. CLASSIFICATION OF ROBOTIZED
CONCEPTS
FOR

CONSTRAWED MOTIONCONTROL

The methods for control of manipulation robots in the constrained motion tasks can be classified according to different
criteria. Considering the kind of the compliance, two basic
groups are distinguished 141, [51.
1) The methods of passive compliance, whereby the real
robots position is approaching the desired position only by
influence of the contact forces themselves.
2) The methods of active compliance, whereby the compliance is provided by using force feedback in order to
achieve either control of the interaction force, or a task-specific
compliance of the robots end-effector.
According to the dominant sources of compliance, the
methods of passive compliance can be classified into the
( 2 ) following two groups.
Xi = Aizi biN(ui) f i P q ; ,
i = 1, 2 , . . . ,n
1) Nonadaptable methods: a) methods based on the inherent
where xi is the (ni x 1) state vector of the ith actuators compliance of the robot mechanical structure, b) methods
model; Ai is the (ni x ni) actuator matrix; bi and f i are which use specially constructed passive deformable devices
the (n; x 1) input distribution and load distribution vectors, (adaptors) attached near the robots end-effectors.
2) Adaptable methods: a) methods based on devices with
respectively; uiis the scalar input to the ith actuator; N ( u i )
is the nonlinearity of the amplitude saturation type; Pqi is the tunable compliance, 2 ) methods based on compliance achieved
load (driving force) acting upon the ith actuator; and ni is by the adjustment of the joint servo gains 161.
Active control force methods may be classified into the folthe order of the actuators state model. The actuator model is
usually of third or second order, where 22 = (42, $, i,T, or lowing two groups: 1) hybrid positiodforce control, whereby
both position and force are controlled in a nonconflicting way
x i = (qa, q i ) T , respectively, and
is the ith rotor current.
In general, the reaction force can be modeled as a complex in two orthogonal subspaces defined in a task specific frame;
fyction of the end-effectors position and motion F = and 2) impedance control, which is in essence based only on
F ( p , 6, j). An accurate environment model, very important the position control and uses different relationships between
for the correct solution of contact tasks in robotics, is usually the acting forces and manipulators position [7].
Taking into account the way in which the force information
difficult to obtain in an analytical form.
is
included in the forward control path, hybrid positiodforce
The first general environment model was proposed in 131.
However, in some practical cases, it is sufficiently accurate to control methods can be classified into the following two
adopt a simplified linearized environment model taking into groups: 1) explicit or force-based methods where force signals
are used to generate the torque inputs for the actuators in
account only the dominant effects:
the robots joints [8]-[10]; and 2) implicit or position-based
algorithms, whereby the force control error is first converted
to an appropriate robots motion adjustment in the forcewhere p is the ( n x 1) vector of the robot external coordinates, controlled directions and then that position is used as input
P E is the ( n x 1) vector of coordinates of the point of in the position controller [11]-[13].
impact between the end-effector (tool) and the environment,
Impedance control methods can be classified into the foland g E ( s ) is the environment model which establishes a linear lowing two groups [14]: 1) position-mode or outer loop
) F . This model may take on control, whereby a target impedance control block relating the
mapping between ( p - p ~ and
one of the following forms:
force exerted on the end-effector and its relative position is
added within an outer control loop for the position controlled
manipulator [12], [15]; and 2 ) force-mode or inner loop
control, where position is measured and force commands are
computed to satisfy target impedance objectives [ 141.
where q is the ( n x 1) vector of the robot joint angles, H , ( q )
is the ( n x n ) inertia matrix, hq(q, 4) is the (n x 1) vector
of centrifugal and Coriolis moments, gn is the ( n x 1) vector
of gravitational moments, Pq is the ( n x 1) vector of driving
forces in joint space, F is the m-dimensional vector of the
generalized forces or of the generalized forces and moments
acting on the end-effector from the environment, and J T ( q )
is the ( n x m ) Jacobian matrix connecting the velocities
of the robots end-effector and the velocities of the robots
generalized coordinates.
In the case of electric dc motors, it is sufficiently accurate
to adopt their dynamic models in the form

ik

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 41, NO. 1 , FEBRUARY 1994

14

METHODS FOR CONSTRAINED MOTION CONTROL

I
I

passive
(methods based on the
passive compliance of
the manipulators
structure)
J

Fig. 1.

active
lmethods where the
compliance is provided
by using force feedback

Basic classification of the methods for constrained motion control

(after PI, 151).

PASSIVE COMPLIANCE METHODS

Fig. 2. Passive compliance classification (after [4], [ 5 ] ) .

(particular directions are


position controlled, and the
remaining directions are force
controlled)

(the mechanical impedance of the robots


end-effector is
controlled1

2 ) Mechanical Compliance Devices: This concept is also


based on the robots structural compliance, but in this case, the
compliance is significantly increased by using specially constructed devices placed at the robots end-effector. Mechanical
compliance devices are implemented especially in robotic assembly. Different types of such devices have been developed,
the best known being the RCC, Remote Center Compliance
[24], developed at the Charles Stark Draper Laboratory. RCC
elements provide additional DOF which increase structural
compliance of the robotic system. The main advantage is
that a simple position controller can be applied, without
any additional force sensor feedback or complex calculations.
However, an RCC element cannot be applied to different
tasks involving parts of different lengths and weights. A
practical solution to this problem is to design a set of different
compliance adaptors which have to be changed according to
the specific tasks. An improvement of RCC is represented
by IRCC, Instrumented Remote Center Compliance, which
provides the fast error absorption characteristic of the RCC
and measurement benefits of a multi-DOF sensor.
B. Adaptable Methods

1 ) Adjustable Mechanical Compliance Devices: Further development of RCC has led to adjustable compliance devices
[25], which enable the location of the center of compliance
Position-based
Force-based
to be mechanically controlled in some prescribed manner,
irpeaance control
inpeaance control
in accordance with parts of different lengths and weights.
Similar to the IRCC, these devices are equipped with sensors
which provide information about endpoint deflections for robot
control.
2) Controller Gain Adjustment: This method is based on
a relatively simple adjustment of position feedback gains in
Fig. 3. Active compliance classification (after [4], [5]).
order to adjust the robots joint stiffness, i.e., in order to get
a desired distribution of the stiffness in different directions
Regarding the force-motion relationship, impedance control of the end-effectors motion. Such adjustment is possible
schemes can be further categorized into: 1) stiffness control
because the end-effectors stiffness depends upon the joint
(F = -KEAz)E)1161, 2) damping control (F = - D E A ~ E ) stiffness, and so it is feasible to calculate the necessary joint
[17]), and 3) general impedance control (F = - ( M E A ~ E
stiffness in order to get the desired end-effectors stiffness. The
DE&E
K E A P E ) )1151, [181-1211.
joint stiffness is determined by the selected position feedback
The above classifications are broadly summarized in Figs.
gain. Therefore, by an appropriate adjustment of the position
1-3. These figures do not include all the known concepts.
feedback gains, we may ensure maximum stiffness in the
In particular, some of the elaborated approaches combine
position-controlled directions and minimum stiffness in the
two or more different methods categorized in distinct groups:
force-controlled directions.
for instance, implicit/explicit control [22], direct compliance
Passive gain adjustment is efficient if the following condicontrol [23], hybrid impedance control [l I], etc.
tions are satisfied: 1) if the static effects (gravitational forces)
are perfectly compensated: 2) if the dynamics of the robotic
IV. PASSIVECOMPLIANCE
METHODS
system can be neglected, i.e., if the robots velocity is low;
3) if the robots tasks are simple, sufficiently specified, and
A. Nonadaptable Methods
previously tested; 4) if the characteristics of the contacted
I ) Structural Compliance: The structural compliance environment are precisely defined; and 5) if this method is
method is based on the inherent robots structural elasticity applied to specific robotic constructions (direct drive robots or
(for example, elasticity of the robots end-effector, elasticity of multifingered hands).
the joints). This method is more important from a theoretical
In reality, application of this method is complicated by
point of view because commercial robotic systems must have the following problems: 1) nonlinear effects such as friction
high positioning accuracy which is achieved by decreasing and backlash in the mechanical transmission, or process phethe inherent robots elasticity properties, i.e., by increasing nomena like jamming, can destroy the stiffness positiodforce
the stiffness of the robots arms.
causality to a great degree; 2) by setting the control gains in

VUKOBRATOVIC? AND TUNESKI: CONTACT CONTROL CONCEPTS IN ROBOTICS

some directions to be very low, the entire system becomes


more sensitive to perturbations; 3) high values of the position
feedback gains in the position-controlled directions may cause
resonant oscillation of the robotic structure; and 4) static forces
such as gravitational forces must be perfectly compensated
because this method does not introduce integral feedback.
All of these facts make the performance of this control
approach uncertain, imposing the necessity for additional
sensor information in order to monitor task execution. Significant improvements of this method are achieved by including
force-sensor information [6], or by introducing an internal
force feedback loop [26]. However, the simplicity of the
passive gain adjustment is lost when these additional strategies
are applied.

15

Fig. 4. Explicit hybrid positiodforce control.

Cartesian space as
U

= - K p S ( p - p) - K,S(1, - $) - KFPS(F- F)

-1

+ K F ~ S ( F - F)Dt
V. ACTIVECOMPLIANCE
METHODS

+ F

(6)

is the ( n x 1) vector of the control inputs ( U =


,u , ) ~ ;K p and K, are the ( n x n) matrices of the
position and velocity feedback gains in the Cartesian space;
K F P and K F I are the ( n x n) matrices of the force feedback
gains in the Cartesian space; and S is the ( n x n ) selectivity
matrix, S = I - S , where 1 is a unit matrix and n denotes
the number of the robots DOFs. Practical realization of the
control method proposed in [8] presents several problems,
which are outlined in Section V-Alc).
concept
b ) Resolved Acceleration Motion Control: This
addresses, in essence, a class of identical methods: the
operational space method [29], and resolved acceleration
control [30]. The proposed control scheme is illustrated in Fig.
5. This method is, in essence, similar to the so-called inverse
dynamics method [31] since, in both methods, the complete
system dynamics is included in the control law. Resolved
acceleration motion control attempts to ensure simultaneous
and nonconflicting position control and force control. The main
aim is to completely decouple the robot dynamic behavior in
the task frame. The control law may be written as
where

(UI, UZ,.

A. Hybrid PositiodForce Control

Hybrid positiodforce control is an active force control


concept based on a theory of compliant force and position
control. The concept was formalized in [27]. The aim of the
hybrid positiodforce method is simultaneous and independent
realization of 1) previously determined nominal trajectories
(positions) of the end-effectors coordinates in the positioncontrolled directions, i.e., in the directions where there is
a natural constraint on the contact force; 2 ) previously determined nominal trajectories of the force components in
the force-controlled directions, i.e., in the directions where
there is a natural constraint on the motion. Each particular
contact task may be defined by introducing an appropriate
task (constrained) coordinate frame, and the so-called natural
and artificial constraints [28]. Natural constraints are a set of
constraints which are a natural consequence of the task configuration. The artificial constraints are defined by the desired
robots motion and/or forces which the end-effector has to
realize. For the sake of simplicity, a diagonal matrix S , called
the compliance selectivity matrix [8], has been introduced
in order to select DOFs which are position controlled and
DOFs which are force controlled (the ith diagonal element is
1 if the ith direction is position controlled, and it is 0 if the
ith direction is force controlled). To specify the constrained
motion, according to [27], the following information sets have
to be defined: 1) position and orientation of the task frame,
2) selectivity matrix S , and 3) desired position and force
expressed in the task frame.
1 ) Explicit Force Control:
a ) Method of Raibert and Craig: The method presented in
[8] is the basic explicit force control method. According to this
method, the control consists of two complementary, parallel
sets of feedback loops (Fig. 4): the upper position feedback
loop, and the lower force feedback loop. Each of these loops
uses separate and independent sensor systems.
Each DOF of the robots end-effector is either position
controlled or force controlled. The main idea of this method is
to ensure cooperative realization of the desired position SX( t )
and desired force ( I - S ) F o ( t )( I is a unit matrix). Following
Fig. 4, the control law proposed in [8] can be written in

Pq = HqJ-l[Sp*- jq]

+ h, + gq + J T T f *

(7)

where Hq is the ( nx n) inertia matrix; J is the ( nx n ) Jacobian


matrix; S is the ( n x n ) selectivity matrix; 3 = I - S , where
I is a unit matrix; h, is the ( n x 1) vector of centrifugal
and Coriolis moments; g, is the ( nx 1)vector of gravitational
moments; p* denotes the inputs from the position control parts
given by
p*

i;

+ K p ( p o - p ) + K,($

-p)

(8)

where p, p , and jF are corresponding nominal position,


velocity, and acceleration, respectively; and f* is the input
from the force control parts, whose form depends on the
applied control law. The nominal acceleration p is also
included in the proposed control law as a feedforward term in
order to compensate for the changes along the nominal motion
in the position-controlled directions.
c ) Drawbacks of Explicit Force Control: The classical hybrid positiodforce control scheme with the explicit force
control has several shortcomings.
First of all, the definition of orthogonal complements,
which is the basis of the explicit hybrid positiodforce control

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 41, NO. 1, FEBRUARY 1994

16

KP

PO

Fig. 5.

Resolved acceleration-motion force control.

scheme, is incorrect [32] because when we use the principle


of orthogonality to determine whether two given instantaneous motions are orthogonal, we encounter: 1) dimensional
inconsistency, 2) dependence on the choice of units used, and
3) dependence on the choice of the coordinate frame. The
definition of natural constraints as orthogonal complements
of artificial constraints is practically meaningless [32]. Using
such a formulation, the subspaces defining natural constraints
are not invariant with a choice of origin or a change of
dimensions.
The main problem with the classical hybrid control concept
is related to the opposite requirements concerning position
and force control subtasks. Namely, position control requires
a relatively high stiffness of the servosystems in the robots
joints in order to ensure precise positioning of the end-effector.
Oppositely, force control requires a relatively low stiffness
of the robot in order to ensure that the end-effector behaves
compliantly with the environment.
In analyzing the formulation of hybrid control [8], [27], in
[33] it has been pointed out that the proposed hybrid control
scheme is not most suitable from the standpoint of practical
realization. Aimed at facilitating the solving of practical tasks,
the authors in [33] have given a reformulation of hybrid control
via the joint space and have introduced orthogonal projection
matrices (calling them filters), which are equivalent to the
selectivity matrix in external coordinates.
According to [34], the following three dynamic problems
have a great influence on the robots stability in the constrained
motion tasks: 1) the effect of the force sensor feedback is
equivalent to high-gain position feedback; 2 ) contact with a
stiff environment is especially critical; and 3) in a robotic
system, due to the sensors and arm elasticity, there is always
an unmodeled high-frequency dynamics which is not taken
into consideration in the mathematical model of the robot in
the constrained motion tasks. These three dynamic stability
problems do not relate especially to the explicit hybrid positiodforce control, but they are already common for many
force control concepts.
Another drawback is that the explicit hybrid positiodforce
control is not robust to task changes and to parameter variation.
It is very difficult to establish some reasonable numerical
or analytical procedure for synthesis of the explicit feedback
gains. Further, the performance of the system depends on the
robots configuration and on the value and the direction of the
desired forces (moments).

Finally, the main problem with the explicit hybrid positiodforce control scheme lies in the fact that it requires a
nonstandard and completely
. new controller. It is not possible
to preserve the classical positional robot controllers which are
very robust and reliable.
The phenomenon called kinematic instability related to
the explicit hybrid positiodforce control scheme will be mentioned in Section VI.
2) Implicit Force Control: The implicit force control scheme
is shown in Fig. 6. This control concept is based on the
identification of the contact stiffness (damping) using the
force-sensor information, and on the computation of the
position (velocity) equivalent to the desired force. The input
to the force controller with transfer function QF (Fig. 6) is
the force error, i.e., the difference between the desired and
the actual contact force in the task frame. The output from
the force controller is an equivalent position ~ E in
Q the forcecontrolled directions which is superimposed on the nominal
position p ( t ) in the position-controlled directions. The sum
( ~ E Q p ( t ) ) is used as a reference input to the position
controller, which remains unchanged. Since the positional
controller provides a basis for realization of the force control,
this concept is referred to as implicit or position-based force
control [12], or external force control [35].
The implicit control law in the Cartesian space can be
written as

= U0 - K,S(p

- p ) - K,S($

- $0) - K p S ( p - PEQ) (9)

where uo is the ( n x 1) vector of the nominal programmed


U~,...,U;)
n~denotes
;
the number of
control (U = (up,
the robots DOFs; Kp is the ( n x n ) matrix of the position
feedback gains in the Cartesian space; K , is the ( nx n ) matrix
of the velocity feedback gains in the Cartesian space; p is the
( n x 1) vector of the end-effectors desired position; p is the
( n x 1) vector of the end-effectors real positions; S denotes
the ( n x n) selectivity matrix, 3 = I - S, where I is a unit
matrix; and ~ E is
Q the ( nx 1)vector of the equivalent positions
in the force-controlled, directions.
The equivalent positions ~ E Qare computed on the basis
of the force controller Q F , inverse environment model g i ,
and the desired force F (Fig. 6). The inverse environment
model ggl is based on the identification of the force model
parameters, and therefore g E 4 i 1 # 1, where QE is the real
environment model (3), (4), or (5). The main problem with
this scheme is the synthesis of the force controller with a
transfer function gF. The role of this element is: 1) to integrate
the force error in such a way that the equivalent position is
proportional to the integral of the force error; in this way,
it is achieved that the steady-state error of force realization
caused by the identification error is reduced by the positional
controller; 2) to compensate the effects of the environment
(process) with transfer function g E [(3), (4), or ( 5 ) ] ; and 3) to
ensure precise following of the desired force trajectory F o ( t ) .
These demands cannot be achieved solely by the positional
controller.
Let us briefly review the effects of the implicit force control
scheme. 1) The basic advantage of this scheme is its reliability
and robustness. This scheme is neither configuration dependent

17

VUKOBRATOVIC AND TUNESKI: CONTACT CONTROL CONCEPTS IN ROBOTICS

COMPUTATION OF

gF AND

KF

IDENTIFICATION
OF 9,

Fig. 6. Implicit force control scheme (after [36])

nor is it sensitive to the parameter variation. The same scheme


may be applied to control various processes. 2) The main
advantage of this scheme over the hybrid positiodforce control
scheme with the explicit force feedback is that, in this scheme,
a classical positional controller in the joint space may be
used. 3) A new control element g F is added in the scheme.
The parameters of the element QF allow enough freedom in
adjusting the system performance. 4) The synthesis of this
scheme is quite clear and straightforward. 5) The proposed
scheme is stable under certain assumptions.
However, this scheme exhibits some drawbacks, too. 1) The
main problem with this scheme lies in the identification of the
contact force characteristics. The system stability highly depends on the identification error. 2) This scheme has relatively
slow responses to force perturbations due to the presence of
the control block g F . 3 ) The accuracy of contact forces is
mainly limited by the precision of robot positioning (sensor
resolution). It can be especially disturbed when contact with a
very stiff environment is required. 4) The performance of the
implicit force control is significantly limited by the bandwidth
of the positional controller. The presence of the force filters
which have to be applied in every scheme also slows down
the response of the system.
3 ) , Implicit/Explicit Force Control Scheme: The implicit/explicit force control scheme is shown in Fig. 7. The
basic idea of this scheme is to combine the implicit and
explicit control schemes in order to include advantages of
both schemes [22], [37]. The advantages of the implicit
control scheme were presented in the previous section, and the
explicit force feedback is introduced in order to: 1) improve
realization of the desired force, 2) speed up the response of
the system to the force perturbations, 3) correct the error
in the identification of the contact force characteristics, 4)
compensate for the effects of friction and force disturbances,
and 5 ) make the system more robust to the errors in guessing
the unknown system parameters.

The implicitlexplicit force control law in the Cartesian space


may be written as
U

= U" - K p S ( p - p " ) - K,S@

-Kps(p -~

p")

+ KFS(F

E Q )

F " ) (10)

where K F is the (n x n) positive definite matrix of the force


feedback gains in the Cartesian space; the other terms have
the same meaning as in (9).
B. Impedance Control

The impedance control is mainly used in contact tasks


where precise positioning is necessary. The impedance control
should ensure: 1) precise positioning; 2) control of the contact
forces; 3) that the end-effector behaves like a system with
corresponding inertial characteristics, i.e., like a system with
desired mass; 4) control of the entire dynamic relation between
the position of the robot's end-effector and the contact forces;
5 ) that the dynamic characteristics of the robot are taken into
consideration; and 6 ) that the prescribed task will be performed
regardless of the parameter variation of the robot, environment,
and regulator.
I ) Position-Based Impedance Control: The position-based
impedance control scheme, proposed in [12], [15], is shown
in Fig. 8. It consists of an inner direct loop from the position
sensor which gives information about the position p~ of the
contact point, and an outer loop based on the force sensor
which gives information about the contact force F . The outer
loop includes a force feedback compensator g F which is
designed in order to ensure stability of the entire robotic
system. The compensator g F defines a relation between the
contact force F and a relative displacement of the end-effector
A ~ FThe
. scheme in Fig. 8 includes transfer functions Q E ,
g p E , and g X E which have the following meaning: gE is the
environmental transfer function matrix (impedance), which in

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS,VOL. 41, NO. 1, FEBRUARY 1994

18

IDENTIFICATION
OF

a,

Fig. 7. Implicit/explicit force control scheme (after [37]).

Dynamic
Compensation

%E

Fig. 8.

Position-based impedance control.

practice may be assumed to be one of the simplified linearized


models (3), (4), or (5); g p E is the control transfer matrix
of the overall robotic system (regulator+robot+environment);
g p E = p/p; p is the vector of the desired external coordinates of the robots end-effector; and g X E is the matrix of
p
describes the influence of
the transfer function p ~ / which
the location of the contact point between the robot and its
environment.
At first sight, the impedance force control scheme (Fig.
8) is similar to the implicit force control scheme (Fig. 6).
However, there are significant differences between these two
schemes: 1) the impedance force control scheme does not
include partitioning into the position- and the force-controlled
directions; and 2) in the impedance control scheme, there is
no prescribed reference force.
The position-based impedance control has the following
positive characteristics: 1) it is robust against the variations
of the parameters of the robotic system and the possible
external disturbances; 2) it ensures reliable working of the
entire robotic system; 3) its implementation is simple because
this scheme does not require a Cartesian positional controller,
but a classical positional controller in the joint space may
also be used; 4) complex calculations are not required; and
5 ) the design of the force compensator gF in the outer loop
is relatively simple. The main drawback of the position-based
impedance control is that it suffers from the inability to provide

Fig. 9. Force-based impedance control.

very soft impedance due to limits in the accuracy of the


positional control system and sensor resolution.
2) Force-Based Impedance Control: The force-based
impedance control scheme, proposed in [14], is shown in
Fig. 9.
In this control concept, an expected reference force F, is
computed to satisfy the desired impedance specification based
on the position error and the inverse of the target impedance
-1

gF :

F, = (ME.?

+ BES+ K E )Ap =

Ap.

(1 1)

The matrices ME, BE, and K E are selected so that the


designed impedance of the robotic system s i 1 corresponds
as much as possible to the real impedance.
The force-based impedance control is reasonable for applications where a small impedance (i.e., small stiffness and
damping) of the robots end-effector QF should be designed,
that is, when a small reaction force F should be performed.
This method is acceptable when the manipulator gravity is
small and slow motion is required. In other cases, the manipulation modeling details are needed.
Similarly to the hybrid positiodforce control, the significant
drawback of the force-based impedance control is that it
requires a nonstandard and completely new controller.

19

VUKOBRATOVI~ AND TUNESKI: CONTACT CONTROL CONCEPTS IN ROBOTICS

VI. STABILITY
ANALYSISOF ROBOTIC
MANIPULATORS
IN CONSTRAINED
MOTIONTASKS

In contact control tasks, three different types of instability


are recognized.
1 ) Dynamic Instability: A simple force control with tip
force sensor feedback is unstable during the contact with stiff
environment, due mainly to the high gain of the wrist force
sensor feedback [34]. From the stability point of view, the
effect of the force sensor feedback is equivalent to very high
gain position feedback.
2) Kinematic Instability: The kinematic instability is a phenomenon related to the explicit hybrid positiodforce control
[38]. It is influenced only by the robots configuration and
the selectivity matrix S. The explicit hybrid positiodforce
control scheme may become unstable in certain manipulator
configurations using revolute joints [39]. However, in [40],
it was shown conclusively that kinematic instability is not
inherent to the hybrid positiodforce control scheme, but is a
result of an incomplete and inappropriate formulation. It was
identified that the inverse of the manipulator Jacobian matrix J
is causing the kinematic instability of the hybrid positiodforce
control scheme.
3 ) Kinestatic Instability: This type of instability is due to
the incompatibility of force/motion command resolution [4 11.
We shall briefly outline all characteristic results in the
stability analysis of robotic manipulators in constrained motion
tasks.
Conventional modeling and analysis techniques were used
for stability investigation of the closed-loop control systems
in [42]. A series of lumped-parameter models were developed
in order to show that a simple force control algorithm exhibits
stable behavior when the higher order dynamics of the robots
arm can be neglected, and it can be unstable if those effects
are significant. The stability of the impliciVexplicit force
control law was partially analyzed in [43]. It was shown
that the stability of the robotic system may be guaranteed
by appropriate selection of the control parameters. The stability analysis of a hybrid positiodforce-controlled robot was
presented in [44]. The authors derived analytical formulas to
estimate accuracy and robustness of the robots at the design
stage. Analytical stability investigations on a discrete two DOF
mechanical model in the case of single-axis implementation
of force control were performed in [45]. The analysis led
to stability charts which are presented on the plane of the
sampling time and force gain parameters. The influence of the
environment stiffness was also analyzed. A force/position local
control scheme of the PID type with gravity compensation was
analyzed in [46]. The scheme is built upon the foundation
of the parallel control strategy which structurally provides
dominance of the force control target over the position one.
Stability is proved via the Lyapunov method. It was shown
that there exists a choice of feedback gains that guarantees
an asymptotic stability of the robot manipulators in contact
with the elastically compliant environment. It was shown in
[47] that in the constrained motion tasks, geometric stability
due to the manipulator configuration and the force-controlled
direction is a significant factor in the overall system stability.

Fig. 10. Experimental environment of debuning.

An analytic and graphic representation of the stable region


which is given for the planar case helps with an explanation
of the intuitive feel of the geometric stability. The stability
properties of the implicit/explicit force control scheme combined with a global control (dynamic compensation) were
investigated for the first time in [48]. Robotic mechanisms
were considered as nonlinear dynamic systems with n degrees
of freedom. The stability is analyzed via the Lyapunov method.
The sufficient conditions to be satisfied by the appropriate
feedback loop gains were derived in order to ensure stability
of the robotic system. An iterative procedure for synthesis of
the feedback gains was given, and it was illustrated by one
practical example.
VII. SIMULATION
AND EXPERIMENTAL
HYBRIDAND IMPEDANCE CONTROL

TESTS OF

To verify hybrid and impedance control modeling, theoretical considerations, and performance analysis, two simulations
and experimental tests have been done. Due to limited space,
only some specific results are presented and briefly discussed.
For the tests, the Manutec R-3 industrial robot was used.
Parameters of the robot and dc actuators which are driving
the joints are given in [49].
A. Implicit Hybrid Control
The force sensor is mounted in the wrist joint measuring
all six components of the force and moment acting upon the
end-effector. The tool for debumng has been attached to the
robot gripper. The parameters of the tool and of the burr are
given in Table I.
The implicit hybrid scheme (Fig. 6) was implemented in
the Advanced Robot Control System ARCOS, which is used
as a control development and test environment at Fraunhofer
Institute, IPK, Berlin, Germany.
The experimental environment is shown in Fig. 10 [50].
The desired cutting force of debumng was 5 [NI. In Fig.
11, real system behavior is presented.

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 41, NO. I, FEBRUARY 1994

20

Time

Fig. 11.

Debuning force versus time.

TABLE I
PARAMETFRS OF DEBURR~NG
PROCESS

Maximal speed of tool

Maximal torque at tool

Stiffness of workpiece

10000

I Nominal width

2000

Nominal depth of burr


force

I"[

I]./.[

of burr

1 Specific cutting

[qm]

0.51cml

0.3[cm]

1500

[N/an]\

Fig. 13.

0.50

0.00.

0.70

Experimental system.

0.90

simulation results

-6.67

Fig. 14. Experimental environment.

higher frequencies in the experiment have appeared due to


sensor oscillations, where the nonlinear friction effects have
a strong influence.

(b)

Fig. 12. (a) Simulation results. (b) Real system behavior,

B. Target Impedance Control

In Fig. 12, the simulation and experimental results of


deburring in the case of high stiffness contact (estimated
stiffness NN 270 000 N/m) with an implicit hybrid control
scheme are presented.
The force and positional bandwidths of the controller are 6
and 8 Hz, respectively.
The technical explanation for the higher experimental bandwidth is that the contact velocity was reached very quickly
(corresponding to impact force). Further, the oscillations with

. . .

The physical system employed in the experiment is depicted


in Fig. 13 [51].
As the working object, i.e., environment, a cantilever beam
with a variable cross section (i.e., stiffness) is used. The
stiffness of the beam is dependent on the contact point location
computed and verified 4 1ieO 1-vukobratovicexperimentally.
The experiments that are reported were performed on the
same manipulator Manutec R-3 robot. The environment
stiffness (cantilever beam) was estimated by force and position

21

VUKOBRATOVICAND TUNESKI: CONTACT CONTROL CONCEPTS IN ROBOTICS

r-----1

------- -' I

Force threshold C
W

Fig. 15. Impedance control scheme.

60

9234

50

40

30

4467
20

10

0
-10

10

-300

5.00

.-.

9.99

Fig. 16. Measured and simulated force during motion.

measurement at several points along the trajectory. The


working trajectory consists of LIN movements between points
A and B (Fig. 14) with a maximal velocity of 20 c d s .
At the start, the robot's end-effector does not contact
the environment (point A), but at the end, it is located
about 8 mm beyond the surface (point B). The environment
stiffness (cantilever beam) was estimated by force and position
measurements at several points along the trajectory. The
stiffness varies from 60 OOO to 400 000 N/m between points
A and B.Based on the global control scheme (Fig. 9), the
impedance control scheme, implemented in the Advanced
Robot Control System ARCOS, is presented in Fig. 15 in a
more concrete form.
The target stiffness Kt = 1500 N/m is selected, and a target
mass of Mt = 20 kg is chosen. In this experiment, the stiffness
of the environment is K E = 157 000 N/m and damping ratio
= 8.

The results of the measured force during motion are presented in Fig. 16. There is a satisfactory match between

measured behavior and simulation results, although the latter


exhibits higher overshoot. The overshoot can be explained by
differences in friction models [52].

VIII. CONCLUDING
DISCUSSION
During the past several years, compliant motion control has
emerged as one of the most attractive and fruitful research
areas in robotics. The control of the constrained motion
of robots is a challenging research area whose successful
solution will considerably affect further application of robots
in industry and increase their efficiency and productivity.
In this paper, we have attempted to present the status in the
area based on work reported in the literature. Although several
control strategies and schemes have recently been proposed
and elaborated, the number of advanced robotic applications
for a complex contact task remain insignificant. The reason
for this is that the majority of new concepts are still in the
laboratory investigation stage, and their implementation into

22

IEEE TRANSACTIONSON INDUSTRIAL ELECTRONICS, VOL. 41, NO. 1, FEBRUARY 1994

todays low-performance control systems is too tedious and


inefficient.
In conclusion, it is perhaps of interest to indicate some
of the subjects for possible future investigation. A clear
formulation and correct solution of positiodforce control are
further required. Simulation and experimental tests of recently
proposed unified approaches to compliant motion control are
also of interest. In impedance control, further advances are
to be expected in the adaptation of target impedances to
complex task requirements. The compliant motion capabilities
analysis of industrial robots and requirements on the next
robot generation from contact task application viewpoints are
of interest to designers. Robust control continues to be in
the focus of control design. Comparison of available algorithms, investigation of compliant control in uncertain and
dynamic environments, solving control problems at higher
control levels, etc., are certainly some areas deserving further
computationaYexperimenta1studies.
In the paper, contact control concepts are classified using
different criteria. Their main characteristics are analyzed separately, and for each of them, essential characteristics are
stated. Different contact concepts are mutually compared,
and their main advantages and drawbacks are outlined. This
review of the contact control concepts covers some early
ideas and their later improvements, as well as recent trends
in this field. The analysis show that two common problems
in the constrained robot motion control are: 1) an adequate
description of the robot-environment dynamics, and 2) a
synthesis of control laws which simultaneously stabilize both
the desired position and interaction force. In order to solve
these problems, the authors of the presented contact concepts
often introduce assumptions which do not hold in reality.
Therefore, due to the specific shortcomings outlined in this
paper, the presented contact concepts have limited applicability
in real contact tasks. Certainly, there are contact control tasks
and specific working regimes where the control concepts,
especially impedance control, may be successfully applied.
Finally, it should be concluded that conventional hybrid
control has some theoretical shortcomings. Weaknesses of this
approach are linked with the notion of orthogonality, and
are caused by the fact that it is incorrect to use the term
orthogonality itself [32]. Finding the directions along which
motion and force are orthogonal, the authors, followers of
the basic concept of hybrid control, use task feedback loops
for stabilization with respect to motion and force separately.
However, this fact does not hold in reality. The attempt
at task simplification by neglecting the friction forces does
not improve the situation considerably since these forces are
unavoidable in many contact tasks in practice (deburring,
grinding, polishing).
It was the reason that the approach, proposed in [53], [54],
is based on the general dynamics equation of the environment
[3], and is simultaneously solving both stabilization tasks of
position and interaction force with the environment. Based
on closed-loop control system stability, the control laws use
both position and force feedback. Furthermore, it is being
shown that in the case of simultaneous stabilization of both
the position and force, the quality of the position transient

response and the force transient response can be controlled.


These tasks are fully solved for conditions of a priori set
constraints on robot motion and robot interaction force with
the environment. Besides that, sufficient conditions are obtained for the stabilization of the desired motion and force
interaction on conditions of external perturbation and errors
of the position and force sensors [SI, and robust force and
position-stabilizing control laws are synthesized [56]. These
initial results, obtained by VukobratoviC: and Ekalo, might be
considered as the contribution to the correct formulation and
solving of this very specific and nonstandard control task in
robotics in a broader sense.

ACKNOWLEDGMENT
The authors have the special pleasure to express their
gratitude to esteemed colleagues, Dr. D. Surdilovic and Dr.
J. Timm, associates of the Fraunhofer Institute-IPK, Berlin,
who in the scope of their successful activity on problems of
contact tasks and their control, have written an excellent survey
report cited in this paper, which has been used by the authors
in parts of the particular formulations and evaluation of the
subject problem, too.
REFERENCES
[I] M. Vukobratovit and V. Potkonjak, Applied Dynamics and CAD of
Manipulation Robots, Scientific Fundamentals of Robotics 6. Berlin:
Springer-Verlag. 1985.
[2] H. Asada and J. Slotine,Robot Analysis and Control. New York Wiley.
1986.
[3] A. De Luca and C. Manes, On the modeling of robots in contact with
a dynamic environment, in Proc. 5th Int. Con$ Adv. Robotics, Fisa,
Italy. 1991, pp. 568-574.
[4] D. Surdilovit and J. Timm, Review of contact control concepts, ESA
Contract 9181/9O/NLIJG(SC), Working Rep. WP 2400, 1991.
[5] M. Vukobratovit and D. Surdilovit, Control of robotic systems in
contact tasks-An overview, Tutorial S5: Force and Contact Control
in Robotic Systems: A Historical Perspective and Current Tec,$nologies,
in Proc. IEEE Int. Conf Robotics Automation, Atlanta, GA, 1993, pp.
13-32.
[6] J. Simons and H. Van Brussel, Force control schemes for robot
assembly, in Int. Trends in Manufacturing Technol.: Robotic Assembly,
IFS Public. Berlin: Springer-Verlag, 1985, pp. 253-267.
[7] D. E. Whitney, Historical perspective and state of the art in robot force
control, Inr. J. Robotic Res., vol. 6, no. 1, pp. 3-7, 1987.
[8] M. H. Raibert and J. J. Craig, Hybrid positiodforce control of manipulators, ASME J. Dynamic Syst.. Meas., Contr., vol. 102, pp. 126-133,
1981.
[9] D. L. Wedel and G. N. Saridis, An experiment in hybrid positiodforce
control of a six DOF revolute manipulator, in Proc. IEEE Int. Con$
Robotics Automation, 1988, pp. 1638-1642.
101 T. Yoshikava, T. Sugie, and M. Tanaka, Dynamic hybrid positiodforce
control of robot manipulators-Controller design and experiment, IEEE
Trans. Robotics Automation, vol. 4, no. 6, pp. 699-705, 1988.
111 R. J. Anderson and M. W. Spong, Hybrid impedance control of robotic
manipulators, in Proc. IEEE Int. Con$ Robotics Automation, Raleigh,
NC, 1987, pp. 1073-1080.
121 J. A. Maples and J. J. Becker, Experiments in force control of robotic
manipulators, in Proc. IEEE Int. Con$ Robotics Automation, 1986, pp.
695-703.
131 J. De Schutter and H. Van Brussel, Compliant robot motion I Formalism for specifying compliant motion tasks, Int. J. Robotic Res., vol. 7,
no. 4, pp. 3-17, 1988.
141 D. A. Lawrence, Impedance control stability properties in common
implementations, in Proc. IEEE Inr. Con$ Robotics Automation, 1988,
pp. 1185-1190.
151 K. H. Kazerooni and B. J. Waibel, Theory and experiment on the
stability of robot compliance control, in Proc. IEEE Inr. Con$ Robotics
Automation, 1988, pp. 71-87.

23

VUKOBRATOVIC AND TUNESKI: CONTACT CONTROL CONCEPTS IN ROBOTICS

J. K. Salisbury, Active stiffness control of a manipulator in Cartesian [44] G. Stepan, A. Steven, and L. Maunder, Theoretical and experimental stability analysis of a hybrid position-force controlled robot, in
coordinates, presented at the 19th IEEE Conf. Decision Contr., Dec.
Proc. 8th CISM-IFIOMM Symp. Theory and Practice of Robots and
1980.
Manipulators, Cracow, Poland, 1990, pp. 53-60.
D. E. Whitney, Force feedback control of manipulator fine motions,
[45] -,
Force control stability and environment compliance in robotics,
ASME J. Dynamic Syst., Meas., Contr., pp. 91-97, 1977.
in Pmc. 8th World Congr. Theory of Machines and Mechanisms,Prague,
N. Hogan, Impedance control: An approach to manipulation, Part
Czechoslovakia, 1991, pp. 503-506.
I-Theory, J, Dynamic Syst., Meus., Contr., vol. 107, pp. 1-7, 1985.
[46] S. Chiaverini and B. Siciliano, On the stability of a fordposition
-,
Impedance control: An approach to manipulation, Part II-Imcontrol scheme for robot manipulators, in Proc. SYROCO Symp., Wien,
plementation, J. Dynamic Syst., Meas., Contr., vol. 107, pp. 8-16,
Germany, 1991, pp. 371-376.
1985.
-,
Impedance control: An approach to manipulation, Part
[47] B. X, I. Walker, D. Tesar, and R. Freeman, Geometric stability in force
control, in Proc. IEEE Znt. Con$ Robotics Automation, San Francisco,
Ill-Application, J. Dynamic Syst., Meas., Contr., vol. 107, pp. 17-24,
CA, 1991, pp, 281-287.
1985.
[48] M. Vukobratovi6 and A. Tuneski, Contributionto the stability analysis
A. A. Goldenberg,Force and impedance control of robot manipulators,
of robotic manipulatorsin contact with environment,Tech. Cybern. (in
IEEE Trans. Robotics Automation, vol. 4, no. 6, pp. 653460, 1987.
D. M. Stokic and D. T. Surdilovic, Simulation and control of robotic
Russian), 1994, to be published.
deburring, Int. J. Robotics Automation, vol. 5 , no. 3, pp. 107-1 15, 1990. [49] M. Otter and S. Turk, The DFVLR models 1 and 2 of the Manutec
Y. Xu, R. P. Paul, and P. I. Corke, Hybrid positiodforce control of
R-3 robot, Rep. DFVLR Inst., Oberpfaffenhofen,Germany, 1988.
robot manipulator with an instrumented compliant wrist, in Proc. 1st
[50] G. Duelen, H. Munch, D. Surdilovic, and J. Timm, Automated force
Int. Symp. Exp. Robotics. Berlin: Springer-Verlag, 1990, pp. 244-271.
control schemes for robotics development and experimentalevaluation,
D. L. Whitney and J. L. Nevins, What is remote centre compliance
in Proc. ZECON 92, San Diego, CA, 1992, pp. 912-918.
D. SurdiloviC and M. VukobratoviC, Impact of target impedance on
(RCC) and what it can do?, Robot Sensors, Vol. 2-Tactile and Noncontact stability, presented at the Int. Conf. Adv. Robotics ICAR93,
Vision, IFS Publ. Berlin: Springer-Verlag, 1986, pp. 3-17.
M. R. Cutkoski and P. K. Wright, Active control of a compliant wrist
Tokyo, Japan.
D. SurdiloviC, S. Anton, and A. AI-Keshmery, Compliant motion
in manufacturing tasks, Robot Sensors, Vol. 2-Tactile and Non-Vision,
control concept for space robotics for upgrading of the ESTEC Robotic
IFS Publ. Berlin: Springer-Verlag, 1985, pp. 17-33.
M. T. Mason and J. K. Salisbury, Robot Hands and the Mechanics of
Laboratory Controller, Study Rep., ESA Contract 103929, Fraunhofer
Manipulation. Cambridge, MA: M.I.T. Press, 1985.
Institute, Berlin, Germany, 1992.
M. T. Mason, Compliance and force control for computer controlled
M. VukobratoviC and Yu. Ekalo, Unified approach to control laws synmanipulators, ZEEE Trans. Syst., Man, Cybern., vol. SMC-11, pp.
thesis for robotic manipulators in contact with dynamic environments,
418432, 1981.
Tutorial S5: Force and Contact Control in Robotic Systems: A Historical
R. P. Paul and B. Shimano, Compliance and control, in Proc. Joint
Perspective and Current Technologies, in Pmc. IEEE Con$ Robotics
Automat. Contr. Con$, San Francisco, CA, 1976, pp. 694-699.
Automation, Atlanta, GA, 1993, pp. 213-229.
0. Khatib, A unified approach for motion and force control of robot
-,
Unified approach to control laws synthesis for robotic manipmanipulators: The operational space formulation,IEEE Trans. Robotics
ulators in contact with dynamic environments, submitted to ASME J.
Automation, vol. RA-3, no. 1, pp. 43-53, 1987.
Dynamic Syst., Meas., Contr.
K. G. Shin and C. P. Lee, Compliant control of robotic manipulators
Yu. Ekalo and M. VukobratoviC, Robust and adaptive positiodforce
with resolved acceleration, in Proc. 24th ZEEE Con$ Decision Contr.,
stabilization of robotic manipulators in contact tasks, Robotica, vol.
Ft. Lauderdale, FL, Dec. 1985, pp. 350-357.
11, no. 4, pp. 373-386, 1993.
R. C. Paul, Modelling, trajectory calculation and servoing of a computer
-,
Stabilizationconditions of robotic manipulators in contact with
controlled arm, A. I. Memo 177, Stanford Art. Intell. Lab., Stanford
dynamic environments,J. Intell. Robotic Syst., 1993.
Univ., 1972.
J. Duffy, The fallacy of modem hybrid control theory that is based on
orthogonal complementsof twist and wrench spaces, J. Robotic Syst.,
vol. 7, no. 2, pp. 139-144, 1990.
H. West and H. Asada, A method for the design of hybrid positiodforce
Miomir Vukobratovit (SM93) was born in Zrencontrollers for manipulators constrained by contact with the environjanin, Yugoslavia, in 1931. He received the B.Sc.
ment, in Proc. IEEE Int. Con$ Robotics Automation, St. Louis, MO,
and PhD. degrees in mechanical engineering from
1985, pp. 251-259.
the University of Belgrade in 1957 and 1964, reC. H. An and J. Hollerbach,Dynamic stability issues in force control of
spectively, and the D.Sc. degree from the Institute
manipulators, in Proc. IEEE Int. Con$ Robotics Automation, Raleigh,
Mashinivedenya, Moscow, in 1972.
NC, 1987, pp. 890-896.
He is presently Director of the Robotics Centre
J. De Schutter and H. Van Brussel, Compliant robot motion 11: A
at Mihailo Pupin Institute, Belgrade, and a Visiting
control approach based on external control loops, J. Robotics Res., vol.
Professor teaching postgraduate courses in robotics
7, no. 4, pp. 17-25, 1988.
at several universities in Yugoslavia and abroad.
D. StokiC, M. VukobratoviC, and D. SurdikoviC, An adaptive hybrid
His interest is in the development of efficient robot
control scheme for manipulation robots with implicit force control, in
modeling and control law synthesis of robotic systems. His special interest
Proc. ICAR 91 5th Int. Con$ Adv. Robotics, vol. 2, Pisa, Italy, 1991,
is dynamic nonadaptive and adaptive control of noncontact and contact tasks
pp. 1505-1508.
in robotics. He is the author or coauthor of 160 scientific papers in the field
D. StokiC, Constrained motion control of manipulation robots-A
of robotics published in leading international journals, and is the author or
contribution, Robotica, vol. 9, pp. 157-163, 1991.
C. H. An and J. Hollerbach, Kinematic stability issues in force control coauthor of 16 monographs published in English, Japanese, Russian, Chinese,
of manipulators,in Proc. IEEE Int. Con$ Robotics Automation, Raleigh, and Serbian.
Dr. VukobratovK is Chairman of the Robotics Section, Yugoslav Society
NC, 1987, pp. 897-903.
H. Zhang, Kinematic stability of robot manipulators under force for Electronics, Telecommunications, Automation, and Nuclear Engineering.
He
is a scientific leader of the national program in robotics, as well as
control, in Proc. IEEE Znt. Con5 Robotics Automation, Scottsdale, AZ,
the principal investigator of three international robotics projects (EC, NSF,
1989, pp. 80-85.
D. W. Fisher and S. M. Mujtaba, Hybrid positiodforce control: A UNIDO). He is also a permanent member of international committees of
several PAC, IFAC/IFIP, and IFToMM symposia, and has for many years
correct formulation, Int. J. Robotic Res., vol. 11, no. 4, pp. 299-311,
been a member of the editorial boards of several leading scientific journals in
1992.
H. Lipkin and J. Duffy, Hybrid twist and wrench control of a robotic robotics, manufacturing,and artificial intelligence. He is a member of ASME
manipulator, Trans. ASME, J. Mechanisms, Transmissions, Automation and the Scientific Society of Serbia, a corresponding member of the Serbian
Academy of Sciences and Arts,and a foreign member of the Soviet Academy
in Design, vol. 110, pp. 138-144, 1988.
S. Eppinger and W. Seering, On dynamic models of robot force of Sciences (now the Russian Academy of Sciences). He was the winner of
control, in Proc. IEEE Int. Con$ Robotics Automation, San Francisco, the highest Yugoslav state award in 1982 for his outstanding achievements in
robotics and technicalcybernetics.He also won, with his coauthors,the highest
CA, 1986, pp. 29-34.
D. Stoki6, Positiodforce control of industrial robots: Implementation scientific Yugoslav award Nikola Tesla in 1986, for their world-recognized
problems, Rep. for Fraunhofer Institute, IPK, Berlin, 1990.
research monograph series published by Springer-Verlag.

.-. .

24

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 41, NO. I , FEBRUARY 1994

Atanasko Tuneski was bom in Prilep, Macedonia,


in 1965. He received the B.Sc. degree in mechanical engineering from the Faculty of Mechanical
Engineering, Skopje, in 1989. Every year during
the fifth-year studies, he won awards as the best
student in his generation. He received the M.Sc.
degree from the University St. Kin1 and Metodij,
Skopje, in 1993. His M.Sc. thesis, Contribution to
the dynamic control of manipulation robots, was
supervised by Prof. M. VukobratoviC.
He is presently an Assistant at the Institute for
Automatics, Hydromechanics, and Pneumatics in the Faculty of Mechanical
Engineering at Skopje. His special interest is dynamic nonadaptive and
adaptive control of manipulation robots in noncontact and contact tasks. He
is beginning work towards the Ph.D. degree, also supervised by Prof. M.
VukobratoviC.

S-ar putea să vă placă și