You are on page 1of 17

On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications

Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
449
ON THE CONTROL OF JOINT INTEGRATED
SERVO ACTUATORS FOR MOBILE
HANDLING AND ROBOTIC APPLICATIONS
Jrg GRABBEL
Technical University of Hamburg-Harburg
Institute for Aircraft Systems Engineering
Nesspriel 5
D-21129 Hamburg
Germany
J.Grabbel@tu-harburg.de
This paper presents a new joint integrated rotary servo actuator design for heavy load mobile manipulators and
robots. The current state of actuator technology in mobile machines is discussed briefly. A new solution for a
pump controlled joint integrated actuator is proposed. For end effector axes this design is ideal for electrical
power distribution, known as power by wire. Two different control strategies are presented and discussed in its
advantages and disadvantages.
Keywords: mobile robot, pump control, joint integrated servo actuator, control strategy, automatic
control
1 INTRODUCTION
The continued progress in automation technology during the past twenty years was clearly
dominated by innovations in the field of stationary robots, e.g. welding and assembling robots
are found in almost every factory. The economic advantages of process automation (decrease
in staff, increase in productivity, automation of monotonous work etc.) are now pushing
towards mobile application. The introduction of more automated machines can currently be
observed e.g. at building sites or cargo and container handling. This is accompanied by a
number of research activities world wide. However, a direct upscale of stationary robot
technology for mobile applications is not possible without any changes. The basic advantage
of electro-mechanical drives and actuators used for stationary systems is founded in the fact,
that electrical energy transmission is considered clean and easy to maintain. On the other
hand, the basic disadvantage of electro-mechanical drives is their comparatively low power
density. This is especially significant in all cases, where actuators are moved with each work
cycle, like nearly all handling and robotic devices do. Since electrical drives of stationary
robots are very large even at comparatively low payloads of 200 kg at max, it is obvious to
On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications
Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
450
use hydraulic drives due to their minimum ten times higher power density for all mobile
devices, where payloads are 10 or 100 times higher than in common manufacturing
applications. Since a Diesel engine is used as primary energy source for all mobile machines
the disadvantage of energy conversion does not weigh too high.
2 STATE OF THE ART
Todays hydraulic drive technology of mobile machines is based almost completely on valve
controlled cylinders. The valves are located next to the driver, who is controlling the machine
by operating the valves directly. High positional precision and high repetitive precision are
not required for todays machines. Also the requirements on the dynamic behaviour are not as
high as they are for stationary robots or for stationary hydraulic systems. Differential
cylinders are preferred due to their more compact design. Currently the constant pressure net
is going to be replaced more and more by load sensing technology. Compared to a pump
working in a constant pressure net, in a load-sensing-system the high pressure pump adapts to
the maximum pressure currently required in the system. Thereby load sensing allows to
operate at least one cylinder at optimal conditions. However, all other cylinders are still most
likely to be operated far away from their optimal conditions. Though, a lot of energy is wasted
to heat due to throttle losses at the control valves. Especially at higher numbers of hydraulic
actuators the energy saving effect of load sensing will not be very high. Additionally load
sensing requires a lot of further hydraulic components compared to a constant pressure net.
Load sensing has also a tendency to positional and pressure oscillations. Furthermore it is not
very suitable for applying automatic control due to its permanently changing plant
parameters, what makes the controller design very complex. However, for automated
processes in mobile handling devices and robots closed loop control of each hydraulic axis is
absolutely essential. Even for half automated processes closed loop control is required, what
will e.g. allow the operator to control tools at the end-effector in Cartesian co-ordinates
instead of controlling each axis angle. Only Cartesian co-ordinate control makes straight line
operation possible, what is required for a lot of processes in automated or half-automated e.g.
at building sites.
3 PUMP CONTROLLED JOINT INTEGRATED SERVO ACTUATORS
A completely new way is the idea of a pump controlled joint integrated servo actuator. Here
the control element, the servo valve, is replaced by a servo pump, what eliminates all
throttling losses within the hydraulic power circuit. Only leakage and friction losses of motor
and servo pump will then reduce the efficiency. Furthermore, the differential cylinders is
On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications
Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
451
replaced by a hydraulic swivel motor in the vane type form. This motor type is very compact
and allows an integration directly into the joint axis. This allow simultaneously a much
simpler design of the joint kinematics, since the original type of motion is a rotation and not a
translation. An additional advantage is that the pressure force is always applied at its optimal
lever arm, what makes the driving torque independent from the joint angle. Also higher angles
up to 270 are achievable by a vane type swivel motor without any disadvantages like long
cylinder strokes. The combination of these characteristics makes this new actuator design
highly suitable especially for end effector joints, where two or three actuators are required to
implement the required number of degree of freedom.
3.1 Principle of operation
The basic principle of the pump controlled joint integrated servo actuator is shown in figure 1.
It consists of a servo pump (1), driven by small electric motor (2), rotating at very high
velocity. The pump is connected to the vane type swivel motor (3) by a closed hydraulic
circuit. The circuit has to two high pressure relief valves (4a and 4b), which release the high
pressure line to the low pressure line in case of overload. An integrated charge pump (5)
supplies the electro-hydraulic servo pump adjustment system (7) and charges the low pressure
line to a typical level of 20 bars to increase the load stiffness. An integrated micro controller
(6) is supplied with positional and/or velocity commands from a central control unit, operated
by the driver and provides all controls for pump adjustment as for velocity and positional
control of the hydraulic axis.
Figure 1: principle of the pump controlled joint integrated servo actuator
Since the joint integrated servo actuator takes its full advantages at the end effector joints, this
also means, that long hydraulic hoses would be needed, when conventional design is assumed,
On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications
Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
452
where the servo pumps are located at a central distribution gear at the diesel engine. This
means also high hydraulic capacities with low stiffness, slowing down actuator dynamics
(response time to position commands). Additionally, long lines increase pressure losses due to
fluid friction and add significant installation mass to the system. A different concept is the
power by wire technology. This technology avoids long lines by transferring the energy
electrically. Instead of using a central power unit or a number of centrally installed pumps,
each actuator uses its own pump, installed locally at actuator. This means the line lengths can
be widely reduced by moving the pump right next to the motor and building a compact unit.
This compact unit brings also some other advantages with it as in case of malfunction or
maintenance this unit can easily be replaced and repaired. The complete testing of these units
can be easily done prior to installation and allows a control parameter adjustment very
comfortable.
3.2 Dynamic characteristics
For analysis of the dynamic behaviour and for closed loop controller design a state space
model of the pump controlled joint actuator needs to be created. It can be derived by using
two basic equations. A complete mathematical model of the actuator was already explained in
Grabbel and Ivantysynova (1999). The continuity equation gives
p&
,
`

.
|


&
2
1
,
M
p Li P
H
V
p k Q
C
, (1)
where C
H
= V/K hydraulic capacity,
V = V
M
+ V
line
oil volume including lines and motor displacement volume.
The terms represent the pump flow (Q
P
), pressure dependent volumetric losses at the
hydraulic pump and motor (coefficient k
Li, p
) and the motor inlet flow due to its movement
(motor displacement volume V
M
). The bulk modulus K is assumed to be constant. The balance
of moments at the servo joint results is calculated to
& & =
{
losses pressure
p
torque friction
R
torque load
load load
torque Motor
M
M M g l m p
V


3 2 1
&
4 4 4 3 4 4 4 2 1
43 42 1
) ( sin
2

. (2)
Here the pressure dependent motor torque, the load (including the mass of the arm itself),
friction effects and pressure losses, which can be subtracted here, are considered.
On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications
Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
453
Looking at the fact, that the servo pump eigenfrequencies are much higher than the
eigenfrequencies of the main circuit, the dynamic behaviour is basically defined by the
eigenfrequencies of the main circuit. Further assuming the load torque as a disturbance,
reducing friction to pure (linear) viscous friction and finally reducing the model to velocity
control, a completely linear state space model of the order two can be derived.
0 ], 1 0 [ ,
0
1
,
2
,

]
]
]
]
]
]

]
]
]
]
]
]

D C B A
H
v
sp
M
H
p Li
C
r
M
V
C
k

with the state vector


]
]
]

&
p
x ,
where pressure losses and fluid mass are also neglected. The third order model for positional
control contributes a third eigenvalue of zero, representing the integrating character of the
hydraulic motor. The load torque has to added, respectively. However, the load torque
function is nonlinear and cannot be added to a single state space model. The following
characteristics can be derived from analyzing the open loop plant of the main hydraulic
circuit:
The system has a dominating, conjugated complex pole pair, which represents the hydraulic
eigenfrequencies of the main hydraulic circuit. This pole pair defines substantially the
dynamic behaviour of the system for applied velocity and positional control. The
eigenfrequencies of the servo pump are significantly higher (10 to 100 times) compared to the
main hydraulic circuit and have no significant impact on the system behaviour. The
dominating pole pair of the main hydraulic circuit is extremely low damped which is a typical
property of all pump controlled systems. This low damping ration is responsible for a
significant oscillations in the main circuit, if no other measures (by means of closed loop
control or added hydraulic components) are implemented. The closed loop system will
become instable already at low gain, when single proportional control is used.
The dynamic behaviour is obviously determined by the poles of the main hydraulic circuit.
Significant parameters for deriving the location of these dominant poles are
for the eigenfrequency
pipe and motor fluid volume and
load torque and inertia;
On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications
Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
454
for the damping ratio
internal leakage and
friction losses.
Especially relevant for the controller design is the fact, that load torque and inertia will
change during operation. What is quite normal for all mobile machines carrying different
loads, makes the controller design difficult, what will discussed later. Since the
eigenfrequencies are dependent on the load torque and inertia, a variation in the load inertia,
will also lead to a variation in the hydraulic eigenfrequencies. Combined with the very low
damping ratio this will basically rule all efforts for an adequate controller design.
3.3 Control strategies
Regarding automated process management and precise, collision free positioning the
following demands have to be fulfilled by designing a suitable control strategy:
overcritical damping ratio for collision free positioning,
high bandwidth for short work cycle duration.
It has to be pointed out, that an overcritical damping ratio is the most important demand, all
other goals are of secondary importance. Since for a significant number of applications
velocity control is required additionally, if the system is subject to trajectory control, where a
drag error is to be minimised, a closed loop velocity control has to be applied as well.
Considering all demands, the controller design can be divided into four major steps:
servo pump swash plate control (not further explained at this stage),
velocity control,
positional control,
increase of system damping.
Considering the swash plate control as integrated into the servo pump a two cascade control
concept is proposed (figure 2). The inner cascade is the velocity control loop, while the outer
cascade is the positional control loop. This kind of axis controller would allow digital
implemention and can be integrated into a power-by-wire- modul as mentioned before.
Positional and velocity commands for axis and trajectory control will than be submitted by
signal wiring from a centralized control unit. For this centralized control unit the power-by-
wire-module appears as a smart actuator unit, what only needs signals of desired position and
maximum velocity.
On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications
Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
455
M
L
= f( )
Q
P
integration
1/s
e
position
controller
velocity
controller
actuator
load torque
e

c

c
Figure 2: two cascade control strategy
3.4 Velocity control
Previous investigations have shown the achievement of overcritical damping to be the most
difficult one, especially if the bandwidth has to be kept as high as possible. For velocity
control a PI-controller has shown to be very effective, where the integrative part is necessary
to compensate control deviation in the velocity control loop, because an integrative part is
missing in the plant. This control structure is shown in figure 3, where for the integrative part
a limited semi-integrator is used.

Arm

Soll
Schw enk -
ar m
Ser vo-
pumpe
-Regler
Q uasi-Int egr at or
D r uck -
r ckfhr . Ts+1
k
figure 3: velocity control using a limited semi-integrator
Investigations have shown that a carefully designed velocity controller, including measure to
increase the damping ratio to desired values, would allow to use simple proportional control
for the positional loop without losing (further) bandwidth.
3.5 Principle of the limited semi-integrator
Looking at other studies done in the field of closed loop control for hydraulic and all other
mechanical systems involving friction, the problem of limit cycles and wind up effects has
always been a considerable problem, where a number of solutions where derived to be more
or less effective, e.g. switching integrators. A new method was developed at the Institute for
Aircraft Systems Engineering (Berg 1999), where the system is prevented from carrying out
On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications
Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
456
limit cycles without any typical integrator problem like wind-ups or defining switching
conditions. This principle, called limited semi-integrator, is shown in figure 5. But first,
consider the transfer function derived from figure 4:
1
) ( ) (
+
+
Ts
k
s E s U , (3)
where k is to be considered as tuning screw and
T is the integrator time constant.
+
+
e
Ts+1
k
u
Figure 4: unlimited semi-integrator
This transfer function can be transformed as follows:
) ( ) ( ) 1 ( ) ( ) 1 ( s U k s E Ts s U Ts + + + (4)
and then ) ( ) 1 ( ) ( ) 1 ( s E Ts s U k Ts + + . (5)
Finally the transfer function yields
T
k
s
T
s
k Ts
Ts
s E
s U
s G
BI

+
+

+
+

1
1
1
1
) (
) (
) ( . (6)
It is now easy to see, that this transfer function has a zero at s
0
= 1/T and a pole at s
P
= (1
k)/T. This means, the semi-integrator consist of pole and of zero, where the position of the
zero depends on T only, while the position of the pole depends on T and k. In case k = 1 this
transfer function can be simplified to
s T s T
s T
s E
s U
s G
I I
I
BI
1
1
1
) (
) (
) ( +
+
(7)
with a zero at s
0
= 1/T and a pole at s
P
= 0. For this case the semi-integrator becomes a true
integrator. In other words reducing k to values from k = 0.95 .. 1.00 would allow to simply
adjust the systems to its friction and stop it from carrying out limit cycles.
On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications
Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
457
Figure 5 shows the same principle, where only a saturation characteristic is added. The
saturation limits the output signal and finally prevents the integrator from running upwards
(wind-up effect).
+
+
e
Ts+1
k
u
Figure 5: limited semi-integrator without wind-up effect
3.6 Measures to increase the damping ratio
As mentioned before the basic disadvantage of pump controlled hydraulic systems is the
extremely low damping ratio of the dominant pole pair of about d = 0.01 .. 0.2. Since
overcritical damping (d > 1) is required to prevent the system from overshooting the damping
has to be increased by suitable measures. As mentioned in the introduction, the design of an
energy efficient actuator was a main reason for the change from valve control to pump
control. That means that a typical hydraulic mean to increase damping, using a bypass
throttle, has to be avoided, since this would also increase the losses. On the other hand, a
number of control measures to increase damping are known. Two of those are discussed here:
Compensation of the low damped pole pair by a complementary pair of zeros
(cancellation), replacing the old pair by a new one with overcritical damping.
Pole placement by stated feedback or reduced state feedback allows free choice of
eigenfrequency and damping ratio, limited only by saturation and eigenfrequency of the
control element.
3.7 Compensation
A good approach to increase the damping ratio is the design of a second order compensator. It
consists of a of complex conjugated pair of zeros and two poles, either complex conjugated or
on the real axis. The purpose of the pair of zeros is to cancel the low damped pole pair of the
plant, while the two poles will replace the cancelled pole pair by forming a new dominant
pole pair. Figure 6 is illustrating this design approach, considering a complex conjugated pole
pair (keep in mind, that the integrator added by the controller is shown here as well since its
dynamic has influence on the pole trajectories). For this case the compensator transfer
function gives
On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications
Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
458
1
1
1
) (
2
2
1
2
2
+ +
+ +

s
d
s
s
d
s
s G
cp
cp
cp
cz
cz
cz
comp

(8)
where
cz
- the compensator zero pair eigenfrequency,

cp
- the compensator pole pair eigenfrequency,
d
cz
- the compensator zero pair damping ratio and
d
pz
- the compensator pole pair damping ratio.
M
L
= f ( )
Q
P
integration
1/s
e
position
controller
velocity
controller
PI
controller
compen-
sator
2nd order
actuator
load torque
e

c

c
Figure 6: block diagram, showing PI-compensator velocity control
The basic advantage of a compensator is its comparatively easy design method, especially for
time discrete implementation state of the art today. Furthermore, this design approach only
needs the positional signal, what needs to be measured anyway. The velocity signal is derived
subsequently by a differential filter. On the other hand, care has to be taken onto the choice of
the eigenfrequencies of the compensating zero pair and poles. It has to be pointed out, that a
complete cancellation of the low damped pole is not possible by a fixed parameter controller
design due to the variation of the plants poles, what is basically provoked by varying loads
(causing varying torque T
load
and inertia
load
). This means, even with closed loop control the
low damped pole will remain, practically, since the compensating pair of zeros is fixed, while
the plants pole pair is moving with varying load. A dominant behaviour of the compensation
pole pair can only be achieved by reducing the compensators eigenfrequency to values below
those of the lowest hydraulic eigenfrequency of the plant. This means explicitly

cp

!
<
hydr., min.
On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications
Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
459
-35 -30 -25 -20 -15 -10 -5 0
-15
-10
-5
0
5
10
15
Root Locus Design
I
m
a
g
A
x
e
s
Real A x i s
do mi nant po l e pai r
o f mai n hydr aul i c ci r cui t
co mpensat o r
zer o pair
int egr at or
zer o
i nt egr at o r
po l e
compensat or
poles

cp

Closed Loop
Figure 7: pole-zero-map of the velocity control loop (PI + Kompensator)
Figure 7 shows the pole zero map of this approach. This also shows the weak spot of this
approach or a any fixed parameter controller design, respectively. As shown above the closed
loop eigenfrequency will also be below the lowest open loop eigenfrequency what causes a
loss of bandwidth especially for low level signal response. This loss of bandwidth can be
quite high, especially for small pays loads, where the open loop eigenfrequency gets very
high. It depends on the dynamic requirements, whether this type of controller design with its
dynamic weakness can be tolerated or not. However, this dynamic weakness concerns
basically the acceleration and deceleration of a procedure, while the total time for large
positional steps is not effected too much.
Figures 8 and 9 show the effect of incomplete pole compensation for different load torque. In
the first case load mass and design mass (the load mass assumed for the controller design)
are almost identical. The second case shows incomplete compensation here the true load
mass is m
load
= 500 kg, while the design mass is m
load, design
= 1500 kg. In both cases the
velocity signal shows overlaid oscillations of the incompletely cancelled pole pair
eigenfrequency. Since the commanded velocity is controlled well within its desired values,
this behaviour may be tolerated in most cases.
On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications
Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
460
Following statements can be made as a conclusion of the compensator design
The dynamic is limited by the lowest frequency of hydraulic pole pair (derived from the
highest pay load). The dominant eigenfrequency of the compensator has to be chosen
significantly below this frequency.
For insufficient compensation the transition behaviour shows underlaid oscillations of the
insufficiently compensated eigenfrequency, what has no effect on the total transition time.
0 1 2 3 4 5
0
5
10
15
20
25
30
Angular position
Time [s]
P
o
s
i
t
i
o
n

[

]
0 1 2 3 4 5
0
5
10
15
Angular velocity
Time [s]
V
e
l
o
c
i
t
y

[

/
s
]
Figure 8: Design for m
load, design
=1500 kg, Step response for m
load
= 1500 kg
0 1 2 3 4 5
0
5
10
15
20
25
30
Angular position
0 1 2 3 4 5
0
5
10
15
Angular velocity
Time [s]
Ve
loc
ity
[/
s]
Figure 9: Design for m
load, design
=1500 kg, Step response for m
load
= 500 kg
On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications
Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
461
Improvements
An improvement can be expected for an adaptive controller adjustment depending on the
current pay load. This would require a load identification method, what can be achieved e.g.
by using pressure sensors. However, only for low and medium pay loads a dynamic
improvement can be expected, since for high pay loads, open loop and closed loop frequency
are quite close to each other. Another idea could be a compensator of higher order, what
would allow eventually to partially eliminate the underlaid oscillations.
4 EXPERIMENTAL RESULTS
For verification of simulation results a test rig was designed to create a real environment for
the joint integrated servo actuator. This test rig consists of an adjustable arm, where the length
and position of the pay load can be changed during operation to simulate varying pay load
mass and inertia during a manoeuvre. This test rig allows the verification of the controller
performance and positional precision in a real situation as well as recording leakage and
friction behaviour of the used hydraulic motors. The angle is measured by TTL angle meter
with a resolution of 0.01. This signal is differentiated to get the velocity signal. The torque at
the motor shaft and pressure in line A and B are measured as well. While the pressure signal
could be used for a future state feedback control concept to increase the damping ratio of the
main hydraulic circuit, the installed torque meter is only needed for result verification.
Figure 10: test rig for experimental verifications
On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications
Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
462
First test rig results have shown the compensator design to work acceptable within its desired
parameters. A step response is shown in figure 11. Here the velocity signal shows a
significant ripple. This has several reasons. One reason can certainly be found in the
incomplete cancellation of the underdamped poles. A second reason, however, is the method
to generate the velocity signal. Since the velocity signal is calculated from the measured
position, the accuracy and time delay of the velocity signal depend on the method used, the
sampling time and the resolution of the angle sensor. For the shown measurements the time
delay of the speed signal is approx. 70 ms, what will consequently lead to a time delay in the
velocity control and to further oscillations. The improvement of the velocity signal calculation
will be one further step of improvement. Future investigations will focus on the optimization
of the compensator design and the implementation of the pressure feedback, a reduced pole
placement design. Implementation of load identification will be considered, if pressure
sensors give satisfactory results.
0 1 2 3 4 5
-5
0
5
10
15
20
25
30
35
Angular position
Time [s]
P
o
s
i
t
i
o
n

[

]
0 1 2 3 4 5
-2
0
2
4
6
8
10
12
14
16
Angular velocity
Time [s]
V
e
l
o
c
i
t
y

[

/
s
]
Figure 11: Test rig results - Design for m
design
= 1500 kg, Step response for m
load
1000 kg
5 CONCLUSION
This paper introduced a new concept for a directly driven joint integrated hydraulic servo
actuators, based on the concepts of displacement control, joint integration and electrical
power distribution (power by wire). This new concept is called joint integrated servo actuator
and combines these three concepts for a number of advantages:
Wide swivel angles (up to 270 with vane type motor),
High efficiency due to pump (displacement) control),
Compact mounting, saving structure and space,
Displacement control allows energy return to other consumers,
On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications
Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
463
Power by wire reduces pipe mass, energy losses due to fluid friction and improves the
dynamic behavior by increasing the hydraulic stiffness.
Especially at the end-effector of mobile handling systems and robots, where these joint drives
can be combined to a compact unit with two or three joint axes, this approach appears
promising. The proposed control concepts appear promising for a fast and easy installation
and initial operation phase. Simulations have shown, that both concepts will be able to
achieve an efficient increase in damping while the overall performance is acceptable.
However, even if pressure feedback is likely to achieve higher bandwidth, compared to a
compensator design, it needs to estimate the current load torque to adapt its pressure
feedback. As long as no load identification is not implemented this will always result in a
certain estimation error, causing the velocity control to either overshoot (if estimation is too
low) or undershoot (if estimation is to high). This makes the pressure feedback design most
likely to gain significant improvement from load identification methods.
Different to control concepts a second major subject to further investigation is the efficiency
and the hereby influenced heat balance in the system. Since an energy efficient actuator
design is desired the loss characteristic of servo pump and motor need to be analyzed. They
determine basically the efficiency and the heat balance in the system. To estimate the
efficiency and the heating of the actuator a mathematical model of the loss behaviour and the
heat transfer has to be developed. This would allow to integrate these system characteristics
into simulation models to derive the efficiency for given work cycles and procedures without
rebuilding each situation with a test rig.
6 LIST OF NOTATIONS
motor shaft angle
arms inertia kgm
Eigenfrequency Hz

cp
Eigenfrequency of compensator Hz

hydr. min
Minimal Eigenfrequency of hydraulic pole pair Hz
p differential pressure at the motor bar
C
H
hydraulic capacity of the pipes from pump to motor m/bar
d damping ratio -
e input of limited semi-integrator
g gravity m/s
k tuning gain of limited semi-integrator -
k
Li, p
coefficient for internal motor leakage l/min/bar
K bulk modulus of compression N/m
On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications
Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
464
l line length m
l
load
distance of arms center of mass to rotational axis m
m
Arm
arms mass kg
m
load
payloads mass kg
M
R
function for friction losses Nm
M
p
function for pressure losses Nm
Q
P
servo pump volume flow l/min
T Time constant of limited semi-integrator s
u output signal of limited semi-integrator -
V
M
motor displacement volume ccm
A System matrix -
B Input matrix (here vector) -
C Output matrix (here vector) -
D Pass through matrix (here scalar) -
x State vector -
7 REFERENCES
Boes, C. 1986. Hydraulische Achsantriebe im digitalen Regelkreis. Dissertation. Aachen:
RWTH Aachen.
Canudas de Wit, C., B. Siciliano, G. B. (Eds., 1996). Theory of Robot Control. Heidelberg:
Springer.
Chiang, M.-H. (1998). Adaptive Achsregelung fr den Hydraulikbagger. 1. Internationales
fluidtechnisches Kolloquium, Aachen.
Eberle, C. 1998. Simulation mobiler Arbeitsmaschinen durch Kopplung der mechanischen
und hydraulischen Teilsysteme. 1. Internationales Fluidtechnisches Kolloquium, Aachen.
Grabbel, J., Ivantysynova, M. (1998). Hydraulic servo joint actuators for mobile
manipulators, 1
st
Bratislavian fluid power symposium, Slovakia.
Grabbel, J.; Ivantysynova M. (1999). Integrated Servo Joint Actuators for Robotic
Applications. 6
th
Scandinavian International Conference on Fluid Power, Tampere, Finland.
Ivantysynova, M.; Grabbel, J. (1998). Hydraulic Joint Servoactuators for Heavy Duty
Manipulators and Robots. 2
nd
Tampere International Conference on Machine Automation
ICMA `98. Tampere, Finland.
On the Control of Joint Integrated Servo Actuators for Mobile Handling and Robotic Applications
Proc. of 1
st
FPNI-PhD Symp. Hamburg 2000, pp. 449-465
465
Ivantysynova, M. (1998). Die Schrgscheibenmaschine eine Verdrngereinheit mit groem
Entwicklungspotential, 1. Internationales fluidtechnisches Kolloquium, Aachen
Ivantysynova, M., O. Kunze, H. Berg (1995). Energy saving hydraulic systems in aircraft
a way to save fuel. 4
th
Scandinavian International Conference on Fluid Power, Tempere.
Mkinen, J., A. Ellman, R. Pich (1997). Dynamic simulation of flexible hydraulic-driven
multibody systems driven using finite strain beam theory. Proc. 5
th
Scandinavian
International Conference of Fluid Power. Linkping/Sweden.
Mar, J.-C., Moulaire. P. (1999). Expert Rules for the Design of Position control of
Eletrohydraulic Actuators. 6
th
Scandinavian International Conference on Fluid Power, May
1999. Tampere, Finland.
Mar, J.-C., Laffite. J.-M. (1995). A Study of the different pole assignment strategies for
position control of hydraulic actuators.. 4
th
Scandinavian International Conference on Fluid
Power, September 1995. Tampere, Finland.
Mar, J.-C. (1993). Synthesis of a high performance electrohydraulic actuator from industrial
components.. 3
th
Scandinavian International Conference on Fluid Power, May 1993.
Linkping, Sweden.
Mattila, J. and T. Virvalo (1997), Computed Force Control of Hydraulic Manipulators. 5
th
Scandinavian International Conference on Fluid Power. Linkping, Sweden.
Roth, J. (1984). Regelungskonzepte fr lagegeregelte elektrohydraulische Servoantriebe.
Dissertation. Aachen: RWTH Aachen.
Saffe, P. (1986). Optimierung servohydraulischer Antriebe fr den Einsatz in
Industrierobotern, PhD thesis, RWTH Aachen.
Westkmpfer, E. and R. Bindel (1998). Mehrgrenregelung hydraulischer Antriebe
Potentiale fr Handhabungsgerte. O+P lhydraulik und Pneumatik, 4/98.

You might also like