Professional Documents
Culture Documents
net/publication/220806797
CITATIONS READS
46 2,389
8 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Peter Lawrence on 12 May 2015.
1
grapple
boom stick
cabin wrist
saw
Z5 Y5
X5
θ5
d5
a4 a5
a1 a2 a3
Z0 Y0 Y4
Y1 Y2 Y3
Z4
X0 X1 X2 X3
θ1 X4
θ2 θ3 θ4
Z3
Figure 3. Excavator Hydraulics Circuit
Z1 Z2
2
tion, forces and torques.
This paper concentrates on the experiments carried
out on the Caterpillar 325FB feller-buncher and in-
cludes 1) a comparison of the two hydraulic systems
experimented with, the rst one employing electrohy-
draulic pilot valves which push the spools of the main
hydraulic valves, the second using variable displace-
ment pumps instead of main valves for each actuating
cylinder, (ii) a discussion of bilateral (force-feedback)
control modalities and their eect on stability and per-
formance.
3
Master Subsystem Stable Slave Subsytem
⊕ -1
Vh
1 Cs
Ve
Fh _ Zm
+ S K v ( Z s + C s+ Z e )
_
Bm
1
Km
S
Ze
Fe
4
Km
6. Coodinated Control Experiments
The most objective comparison of coordinated con-
Kmax trol vs standard joint control would be to have two
( xh > 0 )
identical machines and have an operator use each ma-
Kr chine on identical terrain and identical trees. This
type of study had been previously carried out on a
log loader (see [19]) but was not deemed to be feasible
Knom for a feller-buncher since the same trees cannot be cut
twice. Instead, extended demonstrations were set up,
Kmin
( xh ≤ 0 )
to which machine operators and other representatives
of forest companies, harvesting contractors, suppliers
fe and machine manufaturers were invited to come and
F_max operate the feller-buncher.
Figure 7. Stiness Adjustment Rule For valve control, the machine was delivered to a forest
harvesting site near Kelowna B.C. for a period of one
the cylinder forces by solving simple two-bar and four- week. For pump control, the feller-buncher was moved
bar linkage problems [15, 9]. The end-point wrench to a dierent forest harvesting test site near Kelowna
vector fe can be computed from the arm dynamic B.C. It was also evaluated for one week at a second
equations and a Jacobian transformation as follows: site near Quesnel B.C.
Operator Reaction
J T q fe
() = , D(q)q , C (q; q_)q_ , g(q) (2) Each visitor to any of the sites observed a demonstra-
= , D(q)
q , C (q; q_)q_ , Yg (q)pg
tion, operated the machine for a period of time and
then responded to a questionnaire.
It was observed [20] that that novices could be rea-
where q, J , D, C and g are the joint variables, arm sonably comfortable in controlling the machine after
Jacobian, mass matrix, centrifugal and Coriolis terms several hours compared to much longer times of at
and the gravity force, respectively, and the gravi- least several days to a week for the standard machine
tational terms have been re-written in a "linear-in- controls. A major problem with long learning times is
parameters" form [18]. Since neither the excavator nor the potential damage to the machine during the learn-
the feller-buncher have six degrees of freedom, the Ja- ing period.
cobian in (2) is not invertible. Either a pseudo-inverse Experienced operators observed that the machine was
(smallest norm satisfying (2) or setting components of smoother in operation and somewhat faster - espe-
fe to zero can be used to solve this equation. cially for the pump-controlled version. This was hard
Due to the signicant noise component in the velocity to quantify since operators use dierent manufac-
and acceleration signals q_ and q, and due to the uncer- turer's machines and there was no direct comparison
tainty in the link inertial parameters, only the grav- between machines on the site.
itational terms of the arm rigid body dynamics were Because of the large diameter of the saw and the neces-
used in the above computation. It is essential that the sity of perfectly coordinating the joints to produce a
gravity terms be computed because, without gravity horizontal motion during cutting, it is not uncommon
compensation, machine operation with an active joy- that the butt of the tree will be damaged (split) during
stick would be both tiring and hazardous. The pa- felling using standard controls. It was observed that
rameter vector pg was identied from , Yg (q)pg = 0 that the incidence of butt damage was much reduced
(Yg (q) 2 IR57) by driving the free CAT325FB arm with coordinated control.
very slowly to random locations and collecting the Power and Speed Comparison
cylinder forces and joint variables, then using a recur- During the feller-buncher eld trials, estimates of head
sive least-squares algorithm [15, 9]. With better joint- speed and power consumed by boom, stick and head
angle sensors, the complete forward dynamics could pitch were computed. The bottom trace in Figure 8
be used in the above equation. If the arm parame- shows a typical recording of head speed during valve
ters are not known exactly, the identication proce- controlled operation. The top trace shows the power
dure could be repeated, although there would be sig- consumed by the three functions. Figure 9 shows the
nicantly more parameters to be identied. corresponding recordings during a typical pump con-
5
7. Stiness Feedback Experiments
Summed HP of Boom, Stick, Tilt (HP)
PILOT VALVE CONTROL
150
0
master. The stiness control scheme illustrated in Fig-
−1 ure 6 was implemented along each axis of the hand-
controller, except yaw, with the following parameters
(see Figure 2):
−2
0 5 10 15 20 25
Time (sec)
Y0
150
2.00 4.00 0.06 1.5
100 Z0 3.00 9.00 0.06 1.5
Nm/rad Nm/rad Nm/rad/s mrad
50
X0 -rot. 0.30 0.80 0.04 35.0
0
Y0 -rot. 0.30 0.80 0.04 35.0
0 5 10 15 20 25
Time (sec)
−1
olds were in the range of 6,000 N. End-eector forces
−2
0 5 10
Time (sec)
15 20 25 as high of 55,000 N were encountered.
The overall haptic feedback provided by the above
Figure 9. Machine Power and Head Speed with Pump scheme was excellent. Although a thorough set of mea-
Control surements could not be completed due to the diculty
of working with such a large machine, the ability to
trolled motion. It can be seen from these curves that apply controlled forces was demonstrated. For exam-
higher speeds (conservatively 10-15%) can be achieved ple, Figure 10, displays the result of an experiment
at about 30% lower power. The net benet would be requiring the operator to apply an end-eector force
that machines using pump control would require less against the ground as displayed on the screen of a lap-
cooling, would consume less fuel, and would have a top computer. By contrast, if no stiness feedback
longer working life. There were two disadvantages of is employed, the task becomes impossible to perform.
pump control that were evident. The rst is that the Finally, Figure 12 shows the radial force encountered
hydraulic system is more complex and consequently during tree-cutting.
would cost more than a conventional system. This These experiments have shown that stiness feedback
could be reduced somewhat by improved manifold de- presents a viable way of controlling a large excavator-
sign. The second is a small inherent drift in cylin- based machine in a stable fashion with sucient haptic
der position that has to be actively controlled using feedback to enable the application of controlled forces.
the joint sensors (in valve control, centering the valve Further work is necessary to improve the sensed cylin-
stops cylinder motion). This has safety implications der forces, to quantify the degree of \transparency"
and requires reliable joint or cylinder displacement provided by the stiness feedback, and to demonstrate
sensors. that real tasks can be completed faster when force in-
formation is displayed to the operator.
6
Feller-buncher Control (stiffness control) Feller-buncher Control (without force-reflection)
2 5
joystick
z dead 4
Applied hand force Fh_z(N), Wrist position z(mm)
band
0
3
2
-2
-1
-6
-2
-3
-8
-4
-10 -5
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
Time(seconds) Time(seconds)
5
4
Fz
target
4
target
3
3
Fz
2
2
1
1
0
0
-1
-1
-2 0 5 10 15 20 25 30 35 40
0 5 10 15 20 25 30 35 40
Ti ( d )
Time(seconds)
Figure 10. Controlled Force Application using Sti- Figure 11. Attempts at Controlled Force Application
ness Feedback without Force Feedback
8. Conclusions rick Foster), and the joint angle sensors into the ma-
chine. The late George Olorenshaw of Vancouver de-
Coordinated motion control in either pump or valve signed the hand controller used on the feller buncher.
controlled systems was found by experienced operators Project management was carried out by Rudi Wagner
to provide smoother machine motion and less damage and, arrangements for the eld tests and operator in-
to the tree stems during cutting. It was also found terviews were carried out by Marv Clark of the Forest
that novice operators were able to operate the ma- Engineering Research Institute of Canada.
chine more readily than with the standard controls.
It was found that, although somewhat more complex, Support for the project was derived from the Ad-
a pump controlled machine is both faster and con- vanced Research Institute of B.C., Industry Canada,
sumes lower power than a valve controlled machine. Western Diversication and 6 member companies of
Stiness control was found to be a stable and eective FERIC.
method of controlling force application in endpoint ve-
locity controlled systems. References
As a result of these experiments, we feel that there is a [1] Edwin G. Johnsen and William R. Corliss. Human
considerable potential for improvement in the human Factors Applications in Teleoperator Design and Op-
interfaces in excavator type machines. eration. John Wiley and Sons, New York, New York,
1971.
9. Acknowledgements [2] B.Hannaford, L.Wood, B.Guggisberg, D.McAee,
and H.Zak. Performance evaluation of a six-axis
We would like to acknowledge the following organi- generalized force-re
ecting teleoperator. Technical
zations and persons who greatly contributed to the Report 89-18, California Institute of Technology-
success of this project. RSI Research of Sidney B.C. Pasadena, 1989, Pasadena, California, Jun. 1989.
integrated the pump controlled hydraulic system (Der- [3] Daniel E. Whitney. Resolved motion rate control of
7
Feller-buncher Control (stiffness control) Ibrahim. Ecient computation of the mass matrix
for tree structured systems using orthogonal com-
10
8
plements. In Canadian Conference on Electrical
Wrist position x(mm), Hand applied force Fh_x(N)
4
[9] M. Zhu. Master-slave force-re
ecting resolved motion
control of hydraulic mobile machines. Master's thesis,
x
-6
130 131 132 133 134 135 136 137 138 139 116(2):232{240, Jun. 1994.
Time (seconds)
[11] Alan J. Hewitt. Hydraulic circuit
ow control. U.S.
Patent, allowed Jan 12, 1994, 1994. U.S. ling Jan
Feller-buncher Control (stiffness control)
21, 1993.
4
x 10
2
-3
-5
130 131 132 133 134 135
Time (seconds)
136 137 138 139
Robotics, Kyoto, Japan, Oct 28-30, 1993.
[15] N. Parker. Application of force feedback to heavy
duty hydraulic machines. Master's thesis, University
Figure 12. Hand-Controller and Feller-Buncher Ra- of British Columbia, October 1992.
dial Displacements and Forces During Tree-Cutting [16] M. Zhu and S. Salcudean. Achieving transparency
Process for teleoperator systems under position and rate con-
trol. In Proceedings of the 1995 IEEE/RSJ Interna-
manipulators and human prostheses. IEEE Transac- tional Conference on Intelligent Robots and Systems
tions on Man-Machine Systems, MMS-10(2):47{53, (IROS'95), Pittsburgh, PA, August 5-9 1995.
Jun. 1969. [17] N.R. Parker, S.E. Salcudean, and P.D. Lawrence. Ap-
[4] John M. O'Hara. Telerobotic control of a dex- plication of Force Feedback to Heavy Duty Hydraulic
trous manipulator using master and six-dof hand con- Machines. In Proceedings of the IEEE International
trollers for space assembly and servicing tasks. In Conference on Robotics and Automation, pages 375{
Proceedings of the Human Factors Society-31st An- 381, Atlanta, USA, May 2-6, 1993.
nual Meeting, pages 791{795. Human Factors Society, [18] C.H. An, C.G. Atkeson, and J.M. Hollerbach. Model-
1987. Based Control of a Robot Manipulator. MIT Press,
[5] U. Wallersteiner, P. Stager, and P.D. Lawrence. A Cambridge, MA, 1988.
human factors evaluation of teleoperator hand con- [19] U. Wallersteiner, P. Lawrence, and B. Sauder. A hu-
trols. In International Symposium on Teleoperation man factors evaluation of two dierent machine con-
and Control, pages 291{296, Bristol, England, Jul. trol systems for log loaders. Ergonomics, 36(8):927{
1988. 934, Aug 1993.
[6] Bjorn Lofgren. Kranspetsstyrning. Technical Report [20] Marvin Clark. Coordinated motion of a feller-
Meddelande Nr 18, Skogsarbeten, Box 1184, 16422 buncher. Technical Report TN-223, Forest Engineer-
Kista, Sweden, 1989. ing Research Institute of Canada, Vancouver, B.C.,
[7] M. Ostoja-Starzewski and M. Skibniewski. A master- Canada, 1994.
slave manipulator for excavation and construction
tasks. Robotics and Autonomous Systems, 4:333{337,
1989.
[8] Darrell Wong, Peter D. Lawrence, and Ahmed