You are on page 1of 8

2014 IEEE/RSJ International Conference on

Intelligent Robots and Systems (IROS 2014)


September 14-18, 2014, Chicago, IL, USA

Robust Control of Flexible Joint Robots Based On Motor-side


Dynamics Reshaping using Disturbance Observer (DOB)
Min Jun Kim and Wan Kyun Chung
Abstract This paper proposes a robust control scheme
based on a disturbance observer (DOB) for flexible joint robots.
In this paper, the DOB is applied only on the motor-side
dynamics of the robot, and the uncertainties on the motorside are successfully eliminated. It is shown that, to guarantee
the stability, the estimated motor-side position should be fed
back into the controller, which is different from usual setup of
the DOB. The experiments/simulations show that the estimated
signal feedback indeed guarantees the stability, whereas the
measured signal feedback does not.

I. INTRODUCTION
In recent years, modeling and control of flexible joint
robots (FJR) are gathering more and more interests [1][3].
Flexibility in robot joints originates from flexible mechanisms such as tendon-driven, belt-pulley transmission, harmonic drive as well as the installation of sensors, e.g., joint
torque sensor (JTS). With the diverse growth of the robotics
field, various types of the mechanisms and the sensors are
adopted/developed in robot design, and naturally, control of
FJR is emerged as one of the important topics.
In controlling the FJR, the following model which is
derived under the Spongs assumption [4] is widely adopted.
M (q)
q + C(q, q)
q + g(q) = s

B + s = m + f ,

(1)
(2)

where s = K( q) is measured JTS signal, (1) describes


link-side dynamics and (2) describes motor-side dynamics
with known notations (although f denotes friction torque
in this paper, it may represent any uncertainties, in general).
Noting that the JTS is usually installed after the gearbox,
friction f from which the controller of FJR is significantly
affected is included in the motor-side dynamics and little
uncertainty exists in the link-side dynamics. Performance of
the controller, therefore, is expected to be improved if the
uncertainties in motor-side are successfully compensated. For
this reason, there have been efforts to get rid of friction in
motor-side dynamics.
The most straightforward and classical approach is friction
identification [5]. Another typical approach is motor inertia
reshaping using JTS signals [6], [7]. In this approach, the
inertia B and the friction occurring in the motor-side are
downscaled, so that the motor-side is less affected by the
friction. This approach is proven to be simple and effective,
but does not provides exact reshaping of the motor-side
The authors are with the Robotics Laboratory, School of Mechanical Engineering, Pohang University of Science and Technology
(POSTECH), Pohang, 790-784, Gyung-buk, Korea {mjkim0229,

wkchung}@postech.ac.kr
978-1-4799-6934-0/14/$31.00 2014 IEEE

Controller

Motor-side
Dynamics

Link-side
Dynamics
(nonlinear)

Fig. 1. Motor-side dynamics and link-side dynamics are coupled via joint
torque s . The purpose of this paper is to reshape motor-side dynamics into
the user-defined one using DOB structure.

dynamics because the remaining friction still disturbs the


system, and therefore, inertia reshaping is sometimes used
together with friction identification.
To overcome this, identification-free methods have been
developed [8][12]. Among them, disturbance observer
(DOB) based approaches [8], [11] are particularly interesting in this paper1 ; the DOB, based on the frequency
domain analysis, effectively eliminates the uncertainties, and
thereby the DOB provides exact reshaping of the motorside dynamics. The DOB was first applied to FJR in [8].
However, analysis is provided for only one degree of freedom
(DOF) case rather than the general multi DOF case. The
analysis could not be easily extended to multi DOF case
because of the nonlinear dynamics of link-side (Fig. 1).
On the other hand, [11] proposed a friction observer for
general multi DOF FJRs of which outcome is low pass
filter of the actual friction under the assumption of perfectly
known motor-side model. Moreover, although the observer
has similar structure to the DOB, the stability analysis was
not clear: passivity was proved for the stick-slip friction;
however, the friction is highly nonlinear in general, and
thus, the analysis does not hold on in general. The difficulty
arises from the fact that the motor-side dynamics works
in conjunction with nonlinear link-side dynamics while the
classical identification-free technique is applied only on the
motor-side dynamics (Fig. 1). For this reason, it is difficult
to show the stability of the entire system dynamics.
Therefore, in this paper, it will be shown that the DOB
is actually applicable on the FJR for a class of controllers.
More specifically, we aim at exact reshaping of the motorside dynamics into the user-defined one using DOB, so that
the controller of FJR is no longer affected by the uncertainty
caused in the motor-side. The stability of the entire system
1 the works in [9], [10], [12] also show interesting results, but the
applications are limited to motion controllers, whereas the controller in this
paper can cover force control applications (e.g., impedance control) as well.

2381

Specifically, the exact reshaping of the motor-side dynamics


using the DOB is proposed. To begin with, the following two
assumptions are made2 .
(A1) The controller for FJR (in Fig. 1) satisfies global
asymptotic stability for nominal plant.
(A2) The above global asymptotic stability can be shown
using the following Lyapunov function

Link-side
Dynamics
+

++

+
-

+ -

1
1
V = qT M (q)q + T B
2
2
1

+ ( q)K( q) + Ug (q) + Va (q, q,


, ),
2

(a)

Link-side
Dynamics
+

+
+

++
+

(b)
Fig. 2. (a) Motor-side dynamics reshaping using DOB. (b) An equivalent
structure to the DOB.

where (s) = Pn (s)+[P (s)Pn (s)]Q(s). In low frequency


region (i.e., Q(s) 1), (3) can be reduced to

Link-side
Dynamics
Controller

+
+

+
+

++

where g(q) = (Ug (q)/q)T , and Va > 0 represents any


additional terms.
DOB is a well-known technique that makes the actual
plant behave like a nominal plant, and can be applied on
the motor-side as described in Fig. 2a, where P = P (s)
represents actual plant, Pn = Pn (s) represents nominal
plant, and Q = Q(s) is a low pass filter to be designed.
The transfer function from inputs m s and f to output
is given by
]
[
]
[
P (s)Pn (s){1 Q(s)}
P (s)Pn (s)
(m s ) +
f ,
=
(s)
(s)
(3)

= Pn (s)(m s ).

Fig. 3.
A controller is applied to the FJR with motor-side dynamics
reshaping. Note that is fed back into the controller instead of , and
k is additionally introduced.

In other words, the actual plant P (s) behaves like the


nominal plant Pn (s), and the friction is eliminated. Thus,
design of Q(s) is one of the most important issue in DOB
design. Therefore, lots of studies have been done on the
design of the Q(s) filter, e.g., [18][20].
An alternative construction of the DOB structure is given
in Fig. 2b [21], where C = C(s) is defined by
C(s) =

is shown using the input-to-state stability (ISS) notion. The


control scheme proposed in this paper, however, has one
difference from the usual setup of DOB: estimated signals
(outcome of the nominal plant) should be fed back into the
controller instead of the real (measured) signals. Moreover,
it will also be shown that, in the proposed structure, the 1st
order Q-filter and the 3-2 order Q-filter satisfy a certain H
optimality criterion (i.e., L2 -gain attenuation).
This paper is organized as follows. In section 2, the overall
scheme of the proposed approach is introduced. In section
3, theoretical justification is made. In section 4, validation
of the proposed approach is shown through experiments
and simulations. In section 5, discussion and conclusion are
presented.

Q(s)
.
Pn (s){1 Q(s)}

(5)

By virtue of the alternative structure, the analysis and implementation become simpler as will be investigated in next
section.
Moreover, it should be pointed out that, when the controller is applied, nominal (estimated) signals should be fed
back into the controller instead of real (measured) signals as
shown in Fig. 3. Also, to make the theory clear, k in Fig.
3 should be designed to satisfy

e Tnr s + T k 0.

(6)

One easy choice of k can be (assuming scalar for simplicity)


,
if e nr s 0 and ||

II. C ONTROL OF FJR WITH M OTOR - SIDE DYNAMICS


R ESHAPING

otherwise,

This section summarizes main results of the paper excluding derivations which will be introduced in next section.

(4)

e nr s ,
k = sgn()

k = 0,

(7)
(8)

2 To the authors knowledge, many recent FJR controllers (especially for


the PD based controllers) satisfy the assumptions, e.g., [7], [13][17].

2382

where enr , is some positive constant number


selected to be small if possible. However, in practice, k
can be neglected in many applications (i.e., k 0; will be
discussed more in later sections).
Letting the nominal plant as Pn (s) = 1/(Bn s2 + Dn s),
the following three remarks are worth noticing.
Remark 1 (3-1 order Q(s) filter) [21]: If C(s) is set as PI
control with phase lead compensator of the form
(
)(
)
Ts + 1
1
C(s) = KL
1+
,
(9)
T s + 1
TI s
with KL = Bn /(3 2 ), T = 3 , TI = Bn /Dn , and =
(1 sin())/(1 + sin()) = 1/9, then the following well
known 3-1 order Q(s) filter is obtained
Q(s) =

3( s) + 1
.
( s)2 + 3( s)2 + 3( s) + 1

1
)(s + KP ),
2

(11)

then 2-1 order Q(s) filter is established. If we further set


(K + 12 ) = gBn and KP = Dn /Bn , then the Q(s) is
reduced to the typical 1st order filter as
Q(s) =

g
,
s+g

(12)

where g = B1n (K + 12 ). More interestingly, when PD type


C(s) in (11) is applied, the following H optimality (or
L2 -gain attenuation) is satisfied.

B + s = m + f + k + a

Bn + Dn + s = m + k
M (q)
q + C(q, q)
q + g(q) = s .

(16)
(17)
(18)

Note that the number of equations and states describing


the entire system is increased because the nominal plant
is augmented (e.g., Fig. 3). In the analysis, instead of the
the real plant, we use difference plant that is obtained by
subtracting (16) from (17):
B
enr + De nr = u + w,

(19)

where enr = , u a , and D is damping coefficient


of real plant which may be included in f , and w represents
extended disturbance given by
w = (1 B/Bn )(m s Dn n ) D + f

(20)

Consequently, (17), (18), and (19) describe the entire system


dynamics. For the states describing the entire system, let
q, and
us define xn , xq , and xnr as states related to ,
enr , respectively. The states can be defined in the form

of, for example, xn = [( d )T ( d )T ]T , xq =


[(q qd )T (q qd )T ]T , and xnr = [eTnr e Tnr ]T . Please note
that the definitions of system states depend on the choice of
the controllers and Q(s) filters.
Global asymptotic stability of (17)-(18) will be shown first,
and then the extended disturbance w ISS of (19) will be
shown. The w ISS of the entire dynamics will be shown by
combining these results.
A. Global Asymptotic Stability of (17)-(18)

In this section, the derivations of the results presented in


the previous section are provided. The extended disturbance
w input-to-state stability (ISS) of the entire system is shown.
Let us write the entire dynamics as follows:

(10)

The maximum phase contribution of the lead compensator is


max = 53 at the frequency w = 1/ .
Remark 2 (1st order/2-1 order Q(s) filter): If C(s) is set
as PD control of the form
C(s) = (K +

III. T HEORETICAL J USTIFICATION

(13)

Recalling the assumptions (A1), (A2), we can show the


stability of (1)-(2) using the following Lyapunov function

where xnr [eTnr e Tnr ]T , w represents extended disturbance


(roughly, generalized concept of the disturbance, so that the
friction, modeling uncertainties, etc. can be included) which
will be defined later, S represents state weighting matrix,
and R represents input weighting matrix.

Remark 3 (3-2 order Q(s) filter): If C(s) is set as PID


control of the form

= Vkin (q,
+ Vpot (q, ) + Va (q, q,

V (q, q,
, )
)
, ),
(21)

(xTnr Sxnr + aT Ra )dt 2

wT wdt,

C(s) = (K +

1
1
)(s + KP + KI ),
2
s

(14)

where

Vpotential

The time derivatives of Vkin , Vpot are


= qT (s g(q)) + T (s + m )
V kin (q,
)
V pot (q, ) = ( q)
T K( q) + qT g(q)

then 3-2 order Q(s) filter is established. In this case, the


following optimality criterion is satisfied.

(xTnr Sxnr

aT Ra )dt

= T s qT s + qT g(q).

w wdt,

(15)

where xnr [ eTnr eTnr e Tnr ]T .

1
1 T
q M (q)q + T B
(22)
2
2
1
Vpot (q, ) = ( q)T K( q) + Ug (q).
2
(23)

=
Vkinetic Vkin (q,
)

(24)
(25)

Thus,

2383

+ V a (q, q,

V = T m (q, q,
, )
, ).

(26)

Motor brake
Electric motor

The time derivatives of Vkin , Vpot are


Fig. 4.

+ V a (q, q,
).
)
,
,
V < T m (q, q,

(31)

Here, recall that the nominal signal is fed back into the
).
Noting that (31) has
controller, i.e., m = m (q, q,
,
identical form to (26), and using the invariance principle, the
global asymptotic stability of (17)-(18) can be concluded.
=
For example, consider a PD control of the form m (, )
where d is a desired value and
KC ( d ) KD ,
the gravity g(q) is neglected. Defining Va () = 12 (
= Vkin +
d )T K(d ), the Lyapunov function V (q, q,
, )
Vpot +Va () guarantees the global asymptotic stability of (1)(2). The dynamics (17)-(18) is also globally asymptotically
)
= KC ( d ) KD using
stabilized by m (,

,
s ) = Vkin + Vpot + Va ().

V (q, q,
,
Conversely, since global asymptotic stability of (17)-(18)
is guaranteed, there exists a Lyapunov function Vq,n =
Vq,n (xq , xn ) that satisfies the following by the Lyapunov
converse theorem (theorem 23, and definitions (viii), (x) in
[22]).
Vq,n > 0
Vq,n : radially unbounded
V q,n ||xq ||2 ||xn ||2

(32)
(33)
(34)

B. The Extended disturbance w ISS of (19)


Since the difference system (19) is linear, it is easy to
see that the control input u defined according to (5) always
stabilizes xnr . The state trajectory satisfies
t
eA(t ) Bw( )d,
(35)
xnr (t) = eAt xnr (0) +
0
3 Theoretically, should be designed to ensure (6). In practice, however,
k
e T
nr s is usually overwhelmed by other negative-definite terms, e.g.,
T
D. Therefore, one may choose k 0 in many practical situations.

(36)

where ||e || ke t, k > 0, < 0 and A, B come from


the state-space realization of the closed loop system of (19)
which is guaranteed to be stable (A: Hurwitz). Hence xnr
satisfies the following relation
At

If k is designed to satisfy (6) , Then

1 DOF testbed used in the experimental validation.

||xnr (t)|| ket ||xnr (0)||


t
ke(t ) ||B||||w( )||d,
+

(30)

Belt-pulley mechanism
Harmonic drive

and therefore,

Thus,

(
)


V = e Tnr s + T k T D + T m + V a

Link

Motor-side encoder

,
s ) = Vkin (q,
+ Vpot (q, s ) + Va (q, q,
).

V (q, q,
,
)
,
(27)

= qT (s g(q)) + T (D s + m + k )
V kin (q,
)
(28)
T
T
T
V pot (q, s ) = s q s + q g(q).
(29)

Link-side encoder

Joint torque sensor (JTS)

By the assumptions, investigating (26) and using invariance


principle allow us to conclude the global asymptotic stability
of (1)-(2).
Now, to investigate the stability of the dynamics (17)(18), let us define the following Lyapunov candidate function
with a little bit abuse of notation (Vpot is re-defined as
Vpot (q, s ) = 12 sT K 1 s + Ug (q))

||xnr (t)|| (||xnr (0)||, t) + ( sup ||w||),


0 t

(37)

where (, ) is some class KL function and () is some


class K function. In other words, the difference system
(19) satisfies extended disturbance w ISS4 , and hence the
following ISS-Lyapunov function Vnr = Vnr (xnr ) always
exists.
Vnr > 0
Vnr : radially unbounded

(38)
(39)

V nr x (xnr ) + w (||w||),

(40)

for some class K function x (), and some class K


function w ().
C. The Extended disturbance w ISS of Entire Dynamics
To ensure the extend disturbance w ISS for the entire
system, let us define the ISS-Lyapunov function Ve =
Ve (xq , xn , xnr ) as follows
Ve = Vq,n + Vnr

(41)

Then, it can be easily shown that the entire system is extend


disturbance w ISS combining (32)-(34) and (38)-(40).
IV. S IMULATIONS AND E XPERIMENTS
A. Experiments on 1 DOF Testbed
To verify the proposed approach, experiments were performed on 1 DOF testbed (without gravity) shown in Fig.
4 As a matter of fact, any C(s) stabilizes (19) can be chosen because the
ISS notion is always satisfied for any stabilizing controller C(s) since (19)
is linear. From (5), Q(s) filter is derived as Q(s) = Pn (s)C(s)/{1 +
Pn (s)C(s)}. Particularly, PD/PID control form given in (11)/(14) is a
closed-form solution of H optimal problem [23][25], and therefore
satisfies the optimality criterion (13)/(15).

2384

(a) PD control only.

(b) PD control with inertia reshaping.

(c) PD control with 3-2order DOB.

Fig. 5. PD control only, PD control with inertia reshaping, and PD control with 3-2 order DOB are compared. The performance of the controller is
improved by the exact reshaping of motor-side dynamics using DOB.

4. The FJR controller can be either motion controllers or


force controllers, but this paper used PD control (impedance
control) in verification whose control law is defined by
= KC ( d ) KD ,

m (, )
where KP = 50 N/m, KD = 5 Ns/m represent stiffness
and damping, respectively, and d = 0 is desired motorside position. The controller was implemented on a real-time
OS (RTX) and ran at 500Hz control frequency. It should be
noted that the system significantly suffers from large frictions
caused by belt-pulley mechanism (which is amplified by
160:1 harmonic drive). The nominal dynamics (Pn (s) in
DOB structure) was selected as 1/(Bn s2 + Dn s), where
Bn = 2.3 kg m2 which was obtained by CAD data, and
Dn = 0.01 kg m2 /s which was set arbitrary.
The following three sets of experiments were considered.
1) In the first set of experiments, PD control only, PD
control with inertia reshaping, and PD control with
DOB were compared.
2) In the second set of experiments, DOB was implemented with measured signal () feedback.
3) In the third set of experiments, DOB was implemented
with estimated signal feedback, as proposed in the
paper.
In the second and third sets of experiments, it is verified
feedback is indeed necessary to
that the estimated signal ()
ensure the stability. Experiment scenarios were as follows:
During 5 to 10 sec, external torque was applied (about 15
Nm), and removed. Again, during 20 to 25 sec, external
torque was applied in opposite direction, and then removed.
and
Measured motor position , estimated motor position ,
measured JTS signal are shown in figures. Since the 1DOF
testbed has very high joint stiffness, link-side position is
almost the same as motor-side position, and thus, only the
motor side position is presented in figures.
For the first set of experiments (Fig. 5), when PD control
was applied alone, due to the high friction occurred in the

belt-pulley and harmonic drive gears, the link rarely moved


even though about 15 Nm of external torque was applied
(Fig. 5a; estimated position is not used in the controller,
and is shown only for the comparison; note that is ideal
behavior). The motor did not even return to its original
position due to the friction. When inertia reshaping given
by
m = c + (1 )s ,

(42)

where c is a new control input, and = B/Bn = 4, was applied, the friction and motor inertia were effectively reduced,
and the response showed much better result than the previous
one (Fig. 5b). However, due to the remaining uncompensated
frictions, the motor position could not follow estimated
and moreover, the motor position even could not
position ,
return to its original position. The realized stiffness was about
60 N/m which is larger than the desired value. When 32 order DOB is applied, the motor position followed
well, and the desired stiffness was realized (50 N/m). In
overall, due to the high friction caused in motor-side, the
control performance was significantly degraded. Fortunately,
applying DOB effectively eliminated the friction, and the
performance of the controller was successfully improved.
For the second set of experiments (Fig. 6), DOB was
implemented with measured signal () feedback. When 1st
order Q(s) filter was applied, although could not follow
well, the resulting motion was somewhat plausible (Top
row in Fig. 6a). Nice oscillatory was motion occurred (see,
e.g., 10-15 s), and even returned to its original position.
However, this observation does not mean that the stability
is guaranteed; which is one of the states of the entire
system dynamics does not converge to equilibrium (Bottom
row in Fig. 6a). The resulting plausible motion was due
to, paradoxically, weak compensating ability of 1st order
Q(s) filter: although some amount of enr exists, the 1st
order DOB has no ability to compensate the error. This
observation becomes more clear when higher order Q(s)

2385

(a) 1st order Q(s) filter is applied.

(b) 3-1 order Q(s) filter is applied.

Fig. 6. DOB is implemented with measured state () feedback. Left column: results of 1st order Q(s) filter. Right column: results of 3-1 order Q(s)
Middle row: JTS signal. Bottom row: magnified view of dotted box of the top figure. The
filter. Top row: measured position (), estimated position ().
asymptotic stability is not guaranteed when the measured signal is fed back into the controller.

(a) 1st order Q(s) filter is applied.

(b) 3-1 order Q(s) filter is applied.

feedback. Left column: results of 1st order Q(s) filter. Right column: results of 3-1 order Q(s)
DOB is implemented with estimated state ()
2
Middle row: JTS signal. Bottom row: shows that e nr is dominated by D . P1 = e nr ,
filter. Top row: measured position (), estimated position ().
2

is fed back into the controller, as expected.


P2 = D , and P3 = P1 + P2 . The stability is guaranteed when the estimated signal ()

Fig. 7.

2386

shaft is damped out by the controller (energy dissipation),


and after a while, both motor and link sides will converge to
the equilibriums. However, the oscillation in link-side could
not affect the motor-side motion due to the friction in motorside, and consequently, the oscillation was not damped out.
On the other hand, when the proposed approach was applied
with 3-2 order Q(s) filter, the uncertainties (friction, model
uncertainty) was successfully compensated, and the motion
converged to the desired values (Fig. 9b).
V. D ISCUSSION AND C ONCLUSION
Fig. 8.

2 DOF FJR used in the simulation.

filter was applied (Fig. 6). The 3-1 order DOB made
track better, but even though the friction in motor-side was
compensated well, the asymptotic stability was not satisfied
either (Bottom row in Fig. 6b). Therefore, in overall, it
is shown that feeding back the measured signal does not
guarantee the stability.
For the third set of experiments (Fig. 7), DOB was
feedback; i.e., the
implemented with estimated signal ()
controller has the form of
)
= KC ( d ) KD
m (,
was applied. When 1st order Q(s) filter was adopted,
although steady state error was not zero, the asymptotic
stability was guaranteed (Fig. 7a). The steady state error
can be understood by the fact that the corresponding C(s)
to the 1st order Q(s) filter is PD control input (recall
Remark2). On the other hand, when higher order Q(s) filter
was applied, the performance was significantly improved
(Fig. 7b). Moreover, the bottom rows in Fig. 7a, b show that

e nr was dominated by other negative definite term (D2 )


as discussed previously, i.e., k can be neglected. In overall,
feeding back the estimated signals guarantees the stability of
the entire system, and the performance of the controller can
be effectively enhanced by DOB.
B. Simulations on 2 DOF Planar Arm
To verify the proposed approach for multi DOF robots, this
section presents simulation results on 2 DOF planar robot
described in Fig. 8. PD control with gravity compensation
controller in the form of
= KC ( d ) KD + g(qd ),
m (, )

(43)

with d qd + K 1 g(qd ) was applied (for regulation). The


desired link-side position qd was set as qd = [0.2 0]T
rad. Note that this controller satisfies assumptions (A1), (A2)
[13].
When the controller was applied alone, although the controller (43) is an asymptotically stable, the resulting motion
was not asymptotic stable (Fig. 9a). Interpretation to this
result is as follows: Originally, oscillatory motion in linkside was supposed to be damped out by D gain of the motorside controller; once link motion makes motor move (kinetic
energy is transferred from link to motor), the motion of motor

In FJR model (1), (2), most of uncertainties (denoted


by f ) exist in the motor-side dynamics, and relatively
small uncertainties exist in the link-side dynamics. Therefore,
we can easily expect that the control performance will be
improved if the uncertainties in motor-side can be eliminated
successfully. Considering the FJR as two inertia system
interconnected via spring (and spring force is measured by
s ), it seems possible to apply the DOB on the motor-side
dynamics because all the external force (or torque) exerted
on the motor inertia is known (i.e., m s ), which is typical
situation of DOB applications. One naive (and also typical)
implementation is: applying DOB (e.g., Fig. 2), and feeding
back the measured signals into the controller. However, this
paper shows that the estimated signal should be fed back into
the controller to guarantee the stability of the entire nonlinear
dynamics (Fig. 3).
The proposed scheme is verified by the theoretical analysis
as well as the experiments/simulations. The theory says that
the proposed structure can be applied to a class of controllers,
i.e., the controllers of which global asymptotic stability
can be proved by a certain form of Lyapunov function
for nominal plant. Fortunately, many recent FJR controllers
satisfy this condition. In the experiments/simulations, it is
shown that feeding back the measured motor position signal
() does not guarantee the asymptotic stability. On the other
hand, as expected, the stability is satisfied when the estimated
is fed back into the controller. As a
motor position ()
result, when the proposed structure is applied, the motorside dynamics is exactly reshaped into the user-defined one
by DOB, and thus, the performance of the controller can be
enhanced despite of the uncertainties in motor-side.
ACKNOWLEDGMENT
This work was supported in part by the Engineering
Research Center (No. 2011-0030075), and Global Frontier
R&D Program on Human-centered Interaction for Coexistence (No.2012M3A6A3056423)

2387

R EFERENCES
[1] G. Pratt and M. Williamson, Series elastic actuators, in 1995
IEEE/RSJ International Conference on Intelligent Robots and Systems
(IROS), vol. 1, Aug 1995, pp. 399406 vol.1.
[2] M. Zinn, O. Khatib, B. Roth, and J. K. Salisbury, Playing it safe
[human-friendly robots], Robotics & Automation Magazine, IEEE,
vol. 11, no. 2, pp. 1221, 2004.
[3] A. Albu-Schaffer, O. Eiberger, M. Grebenstein, S. Haddadin, C. Ott,
T. Wimbock, S. Wolf, and G. Hirzinger, Soft robotics, Robotics &
Automation Magazine, IEEE, vol. 15, no. 3, pp. 2030, 2008.

(a) PD+gravity controller is applied without DOB

(b) PD+gravity controller is applied with 3-2 order DOB

of the 1st joint. Middle row: those of


Fig. 9. Simulation on the 2 DOF planar arm under gravity. Top row: measured position (), estimated position ()
the 2nd joint. Bottom row: JTS signals. The result shows that the performance of the controller is successfully improved by DOB.

[4] M. W. Spong, Modeling and control of elastic joint robots, ASME


J. Dyn. Syst. Meas. Control, vol. 109, pp. 310319, 1990.
[5] A. Albu-Schaffer and G. Hirzinger, Parameter identification and
passivity based joint control for a 7 dof torque controlled light weight
robot, in IEEE International Conference on Robotics and Automation
(ICRA), vol. 3, 2001, pp. 28522858.
[6] A. Albu-Schaffer, C. Ott, and G. Hirzinger, A unified passivitybased control framework for position, torque and impedance control of
flexible joint robots, The International Journal of Robotics Research,
vol. 26, no. 1, pp. 2339, 2007.
[7] A. Kugi, C. Ott, A. Albu-Schaffer, and G. Hirzinger, On the passivitybased impedance control of flexible joint robots, Robotics, IEEE
Transactions on, vol. 24, no. 2, pp. 416429, 2008.
[8] K. Kaneko, S. Kondo, and K. Ohnishi, A motion control of flexible
joint based on velocity estimation, in 16th IEEE Annual Conference
of Industrial Electronics Society, 1990, pp. 279284.
[9] A. Hace, K. Jezernik, and A. Sabanovic, Smc with disturbance
observer for a linear belt drive, Industrial Electronics, IEEE Transactions on, vol. 54, no. 6, pp. 34023412, 2007.
[10] J. S. Bang, H. Shim, S. K. Park, and J. H. Seo, Robust tracking and
vibration suppression for a two-inertia system by combining backstepping approach with disturbance observer, Industrial Electronics,
IEEE Transactions on, vol. 57, no. 9, pp. 31973206, 2010.
[11] L. Le Tien, A. Albu-Schaffer, A. De Luca, and G. Hirzinger, Friction
observer and compensation for control of robots with joint torque
measurement, in IEEE/RSJ International Conference on Intelligent
Robots and Systems, 2008, pp. 37893795.
[12] S. E. Talole, J. P. Kolhe, and S. B. Phadke, Extended-state-observerbased control of flexible-joint system with experimental validation,
Industrial Electronics, IEEE Transactions on, vol. 57, no. 4, pp. 1411
1419, 2010.
[13] P. Tomei, A simple pd controller for robots with elastic joints,
Automatic Control, IEEE Transactions on, vol. 36, no. 10, pp. 1208
1213, 1991.
[14] C. Ott, A. Albu-Schaffer, A. Kugi, S. Stamigioli, and G. Hirzinger,
A passivity based cartesian impedance controller for flexible joint
robots-part i: Torque feedback and gravity compensation, in IEEE
International Conference on Robotics and Automation (ICRA), vol. 3,
2004, pp. 26592665.

[15] A. De Luca, E. Guglielmelli, and P. Dario, Compliance control for an


anthropomorphic robot with elastic joints: Theory and experiments,
Journal of dynamic systems, measurement, and control, vol. 127, p.
321, 2005.
[16] A. De Luca, B. Siciliano, and L. Zollo, Pd control with on-line gravity
compensation for robots with elastic joints: Theory and experiments,
Automatica, vol. 41, no. 10, pp. 18091819, 2005.
[17] L. Le Tien, A. Schaffer, and G. Hirzinger, Mimo state feedback
controller for a flexible joint robot with strong joint coupling, in
IEEE International Conference on Robotics and Automation (ICRA),
2007, pp. 38243830.
[18] K. Ohnishi, A new servo method in mechatronics, Trans. Jpn. Soc.
Elect. Eng, vol. 107, pp. 8386, 1987.
[19] K. Yamada, S. Komada, M. Ishida, and T. Hori, Characteristics of
servo system using high order disturbance observer, in Proceedings
of the 35th IEEE Decision and Control, vol. 3. IEEE, 1996, pp.
32523257.
[20] Y. Choi, K. Yang, W. K. Chung, H. R. Kim, and I. H. Suh, On the
robustness and performance of disturbance observers for second-order
systems, Automatic Control, IEEE Transactions on, vol. 48, no. 2,
pp. 315320, 2003.
[21] B. K. Kim, H.-T. Choi, W. K. Chung, and I. H. Suh, Analysis and
design of robust motion controllers in the unified framework, Trans.
ASME, J. Dyn. Syst., Meas., Contr., vol. 124, no. 2, pp. 313320, 2002.
[22] J. L. Massera, Contributions to stability theory, The Annals of
Mathematics, vol. 64, no. 1, pp. 182206, 1956.
[23] Y. Choi, W. K. Chung, and I. H. Suh, Performance and H
optimality of pid trajectory tracking controller for lagrangian systems,
Robotics and Automation, IEEE Transactions on, vol. 17, no. 6, pp.
857869, 2001.
[24] M. J. Kim, S. Park, and W. K. Chung, Nonlinear robust internal loop
compensator for robust control of robotic manipulators, in IEEE/RSJ
International Conference on Intelligent Robots and Systems (IROS),
2012, pp. 27422748.
[25] M. J. Kim and W. K. Chung, Design of nonlinear h optimal
impedance controllers, in IEEE/RSJ International Conference on
Intelligent Robots and Systems (IROS), 2013, pp. 19721978.

2388

You might also like