H4 Parallel Robot: Modeling, Design and Preliminary Experiments
H4 Parallel Robot: Modeling, Design and Preliminary Experiments
H4 Parallel Robot: Modeling, Design and Preliminary Experiments
Abstract. This paper first recalls the concept of H4 only 5 controlled axis -plus the spindle rotation-) and
parallel mechanism; the kinematic models necessary for from their limited tilting angle. Solutions to these
real-time control are derived. A simple and efficient drawbacks have been proposed such as the smart -but
control system based on a combination of Windows NT complex- Eclipse machine [ 1 I], or the hybrid parallel-
and RTX is used to demonstrate its performances. Results serial Tricept. We believe that another approach could be
prove the efficiency of H4 serving as a high-speed pick- valid too: one can provide the spindle with 4 axis and let
and-place robot. the work piece be moved by an additional 5‘h axis,
following the left-handright-hand robotics paradigm.
It is amazing to remark that only few efforts have been
1. Introduction
devoted in the past to 4 dof parallel mechanisms. Apart
In the late SO’S, after the first ideas of parallel from Koevermans flight simulator [ 121 and Reboulet 4-
mechanisms proposed by Cough [ l ] or Stewart [2], dof wrist [13], which both provide 3 rotations and 1
Clavel proposed the famous Delta structure [3] as a base translation, we can mention few hybrid (that is to say, non
for a “family” of parallel machines dedicated to high- fully-parallel) mechanisms as in [14] or [15].
speed applications. As far as industrial use is concerned In the following sections we recall the concept of H4
the Delta robot is definitely a success, and this is not so mechanism that was introduced in [ 161; we describe the
common among the large number of different structures detailed design of one possible implementation of H4
proposed by Academic researchers in the last 20 years. concept and briefly derive its kinematic modeling. Our
Indeed, only the so-called “hexapod” built with 6 U-P-S control implementation, based on a combination of
chains in parallel (U-P-S: Universal-Prismatic-Spherical) Windows NT and a Real-Time Kernel, RTX, is then
has been also used intensively in industry (see [4] and [5] described and preliminary results show its effectiveness.
for an extensive coverage of this issue). This may be seen
as a result of, on one hand, the exceptional simplicity of
2. H4Concept
the Delta 3-dof solution, and on the other hand, the
enormous research effort dedicated to the so-called In [I61 we proposed to build a fully-parallel mechanism
“hexapod”. with no passive chain, able to provide high performances
Of course, many alternate designs have been proposed in terms of speed and acceleration. Those considerations
(some of them close to the Delta family, e.g. the machine lead to three important consequences: the mechanism is
proposed by Toyoda, HexaM [6] which is an evolution of based on 4 independent chains between the base and the
Hexa robot [7]). But it is clear today that most efforts are nacelle; each chain is actuated; each actuator is fixed on
dedicated to 6 dof machines which are now well known the base. We did so because such ideas have already
(see [8] for an exhaustive enumeration) or 3 dof machines proven their efficiency for high-speed equipments like the
(see [9] and [IO] for good examples of such devices). Delta robot, the Hexa robot and the HexaM machine-tool.
However, we strongly believe that there is a need for
equipment providing more than 3 dof arranged in parallel
and based on simpler arrangements than 6-dof structures.
As a matter of fact, for most pick-and-place applications,
at least four dof are required (3 translations and 1 rotation
to arrange the object in its final location). On the Delta
robot, this is achieved thanks to an additional U-P-U link
between the base and the gripper. In some cases, this
solution is smart and elegant. In those other cases
(namely, for Delta with huge workspace, or even more,
for linear Delta), this does not seem as efficient as a fully
parallel arrangement.
On the other hand, 6-dof fully-parallel machines currently
in use in machining suffer from their complexity (they Figure 1. The basic idea of H4 concept.
need at least 6 motors while the cutting process requires
2001 IEEE
0-7803-6475-9/01/$10.000 3256
The basic idea of H4 concept is described by a simple We have also shown that the two passive revolute joints
architectural scheme (Figure 1) where joints’ are located originally on the nacelle can be grouped together
represented by rectangles2, and links between those joints with other joints to lead to the architectural scheme of
are represented by lines. Each kinematic chain is a P-U-U Figure 4, called “Asymmetrical H4”, which makes use of
or a R-U-U chain that must satisfy geometrical conditions two Delta-like chains and two Hexa-like chains. Again,
to guarantee that the mechanism offers 3 translations and the P-U-U chains can be replaced by P-(S-S)l chains as
1 rotation about a given axis (see [16] for details). We shown in Figure 5 and Figure 6.
have shown in [ 161 that this concept is indeed a complete
family of 4-dof parallel mechanisms: the P-U-U
(respcctively R-U-U) chains can be replaced by P-(U-S)z
or P-(S-S)2 chains (respectively by R-(U-S)* or R-(S-S)*) -
as shown in Figure 2 and Figure 3 .
--
’ P: prismatic; R: revolute; U: universal; S: spherical. Figure 6. H4 mechanism with 4 rotary drives and an
Grey rectangles represent actuated joints. asymmetrical design.
3257
3. Detailed Design
In this paper we focus on one particular design, similar to
the one depicted in Figure 3. The practical design is
extremely simple thanks to the use of DD motors3
providing a very good position resolution4. Rods are made
of carbon fiber; arms, forearms and nacelle are made of U -
aluminum alloy.
4. Kinematic Modeling
In this section we derive relationships between actuator's
and nacelle's positions represented by, respectively,
q = [ q ,, q 2 , q 3 , q 4 I fand x = [x,y,z,e:l' (in @,U, , u Y ,U, N.
Figure 7. The H4 prototype. We also present the relationship between actuator's and
nacelle's velocities represented by q and X ,
Figure 7 is a photography of our H4 prototype. To extend
respectively, since we need it for the forward kinematic
the range of motion of the nacelle (initially limited to _+ 45
model.
degrees) and to reach a 180-degree rotation capability in
both directions, we have improved the nacelle design and motor
I forearm
equipped it with a mechanical amplification system
shown in Figure 8. A gear system (with a ratio of 4:l)
amplifies the original motion of the nacelle. The complete
architecture is described by Figure 9 where the circle-and-
arrow pictogram indicates the dependency between
motions of two R joints.
3258
where: Applying (3) to the four chains leads to:
M, = -21 cosa, +tA,,. sincx,) A,B, *U, A,B, *U? A,B, *U, (DC, xAIBI)*u,
N, = 21 I?A,z A2B, U, A2B2 ug A2B2 U, (DC, xA2B,) U,
Gi = L2-I2-piAiZ A3B3 U, A3B3 uy A3B3 U, (DC, x A3B3) U,
A4B4 U, A4B4 uy A4B4 U, (DC, xA4B4) U,
Resorting to the following new variable:
'i
t i = tan-
2
we have:
-b, *Jb;Z-4a; c;
q i = 2 tan-'(
2 a,
1
where: J, = diag((PiBixAiBi)mui)
3259
This is a G.U.I. that let us set displacement parameters The mailbox is actually a memory shared by real-time and
(desired position, maximal velocity . . .) and control non-real-time processes. It allows the user's interface to
parameters. User can query data as well. place a request on the mailbox and the control process to
send a response. A simple communication protocol is
User implemented and includes functions to send an order, to
Commands wait for a response and to reply to an order.
By its nature, this control organization is 100-percent
t open and is ready for further improvements.
MailBox Position
6. Preliminary Results
(Shared Control
(WinNT) Memory) This section presents preliminary results obtained for the
following typical displacement:
~ e a and
l desired actuators position Real and desired nacelle s position
Figure 12. The complete control runs on a single PC.
Real-time process. 7
50
motor 4
motor 3
0 4
0.31-
~
1
nacelles angle
motor 2
0.2
lop'
O!->
0 100 200 300
-0 5 i
0
I
i-
100
.-.__.___
200
~~~
300
Tme (ms) Time (ms)
I MAILBOX
3260
References
I 0.03!
[I] Gough V.E., Contribution to discussion of papers on research in
automotive stability, control and tyre performance. Proc. Auto Div.
Inst. Mechanical Engineers, 1956-1957.
[2] Stewart D., A platform with 6 degrees of freedom. Proc. of the
Inst. ofMech. engineers, lZO(Part 1, 15), pp. 371-386, 1965.
[3] Clavel R., Une nouvelle stmcture de manipulateur parallele
pour la robotique Ikgere, APII, 23(6), pp. 501-519, 1989.
[4] Merlet J.-P., Les robots paralleles, Yd Edition, Hemes, 1997.
[5] Tonshoff H.K., A systematic comparison of parallel kinematics.
Keynote in Proceedings of the First Forum on Parallel Kinematic
Machines, Milan, Italy, August 31- September 1, 1998.
[6] Pierrot F., Shibukawa T., From Hexa to HexaM. In Proc.
IPK98: lntemationales Parallelkinematik-Kolloquium, Zurich,
June 4, 1998, pp. 75-84.
[7] Pierrot F., Dauchez P. and Foumier A., Fast parallel robots.
lime (ms) Tme (ms)
Journal of Robotic Systetns, 8(6), pp. 829-840, 1991.
[SI Innocenti C. and Parenti-Castelli V., Exhaustive enumeration of
Figure 15. Tracking errors. fully parallel kinematic chains. Dynamic System and Control, Vol.
55-2, pp. 1135-1 141.
During displacement, nacelle's maximal velocity is close [9] Gosselin C.M. et al., On the direct kinematics of general
to 3 m/s and maximal acceleration is 5 g (Figure 16). A spherical 3-degree-of-freedom parallel manipulators. In ASME
Biennial Mechanisms Conference Proc., Scottsdale, Arizona, pp.
.avi movie can be downloaded from our web site [18]. 7-11.
[IO] Tsai L.W., Kinematics of a three-dof platform with three
extensible limbs. In Recent Advances in Robot Kinematics, pp.
401-410, Kluwer, 1996.
[I I] Kim J. and Park F.C., Eclipse - A new parallel mechanism
prototype. Position paper in Proceedings of the First European-
American Forum on Parallel Kinematic Machines, Milan, Italy,
August 31- September 1, 1998.
[I21 Koevermans W.P. et al., Design and performance of the four
dof motion system of the NLR research flight simulator. In
AGARD Conf. Proc. No 198, Flight Simulation, La Haye, 20-23
October 1975,pp. 17-1117-11.
[ I31 Reboulet C. et al., Rapport d'avancement projet VAP, theme
401
\
.
I ,
7, phase 3. Rapport de Recherche 7743, CNESIDERA, January
0. ,A:i
1991.
.,.-. .so!-.--- -..., ~. ~~~ ~
[14] Hesselbach J., Plitea N., Frindt M., et Kusiek A., A new
parallel mechanism to use for cutting convex glass panels. In
ARK, Strobl, June 29-July 4, 1998, pp. 165-174.
[IS] Tanev T.K., Forward displacement analysis of a three legged
four-degree-of-freedom parallel manipulator. In ARK, Strobl,
June 29-July 41998, pp. 147-154.
[I61 Company 0. and Pierrot F., A new 3T-IR parallel robot,
ICAR '99, Tokyo, Japan, October 25-27, 1999, pp. 557-562.
7. Conclusion [ I71 httu:i/www.vci.comiuroductslvci uroductsirtxlrtx index.html
[ 181 http://www.lirmm.fri-pierrot!
In this paper, after recalling the concept of H4, we have
derived thc models necessary for its control in real-time.
We have shown it is possible and efficient to control such
a high-speed mechanism thanks to a single PC. This is Acknowledgement
made feasible by resorting to the combination of The authors would like to thanks Prof. Reg Dunlop, from
WindowsNT (for the easiness to develop the G.U.I. and Canterbury University in Christchuch, New Zealand, for
its efficient programming tools) and RTX (for the real- his highly valuable remarks regarding mechanical
time facilities). A simple PID controller in the joint improvements of our prototype.
position loop (together with a PI controller in the velocity
loop) leads to nacelle's accelerations of 5g.
With such preliminary results, the H4 robot, which is
based on proven technologies coming from Delta and
Hexa robots, has demonstrated its ability to serve as an
efficient pick-and-place robot.
Further developments will be conducted towards the
direction of more demanding applications such as
machining, where we expect H4 to be part of an hybrid
mechanism able to do 5-axis machining at high speed.
3261