Sie sind auf Seite 1von 11

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/328707069

Workspace Analysis and the Effect of Geometric Parameters for Parallel


Mechanisms of the N-UU Class

Conference Paper · August 2018


DOI: 10.1115/DETC2018-85258

CITATIONS READS
3 375

3 authors, including:

Divya Shah Alberto Parmiggiani


Istituto Italiano di Tecnologia Istituto Italiano di Tecnologia
6 PUBLICATIONS   10 CITATIONS    33 PUBLICATIONS   314 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

R1 humanoid robot View project

Robot Design for Dexterous Manipulation View project

All content following this page was uploaded by Divya Shah on 18 February 2019.

The user has requested enhancement of the downloaded file.


Proceedings of the ASME 2018 International Design Engineering
Technical Conferences and Computers and Information in Engineering Conference
IDETC/CIE 2018
August 26-29, 2018, Quebec City, Quebec, Canada

DETC2018-85258

WORKSPACE ANALYSIS AND THE EFFECT OF GEOMETRIC PARAMETERS FOR


PARALLEL MECHANISMS OF THE N-UU CLASS

Divya Shah ∗ Giorgio Metta Alberto Parmiggiani


Università degli Studi di Genova iCub Facility, iCub Facility,
& iCub Facility, Istituto Italiano di Tecnologia Istituto Italiano di Tecnologia
Istituto Italiano di Tecnologia Genova, Via Morego 30, 16163. Genova, Via Morego 30, 16163.
Genova, Italy Italy Italy
Email: divya.shah@iit.it giorgio.metta@iit.it alberto.parmiggiani@iit.it

ABSTRACT l1 , l2 , l3 Linear geometric parameters


N-UU class mechanisms, exemplified by the Omni-Wrist III, θj Joint angles
are compact parallel kinematic mechanisms (PKM) with large x, y, z Platform Cartesian coordinates
singularity free workspaces. These characteristics make them n, o, a Platform roll, yaw, pitch angles
ideal for applications in robot wrists. This article presents the J Mechanism Jacobian matrix 2x2
detailed kinematic and workspace analysis for four N-UU class
mechanisms. More in detail, the equations defining the mecha-
nism’s moving platform kinematics are derived as a function of
INTRODUCTION
the motion of the input links; these are then used to explore the
With the modern research in the field of humanoids focused
mechanism’s workspace. These results are furthermore validated
more on it’s applications in social situations and human-robot
by comparing them to the results obtained from CAD-based sim-
collaborations, it is essential for the robots to match the dexter-
ulations.
ity of their biological counterparts in grasping objects. These
The analyses suggests that the workspace of the mechanism
skills form an important part of their interaction with environ-
is non-uniform, with a “warping” behaviour that occurs in an
ment and their capacities for exploring, feeling, learning, plan-
asymmetric fashion in a specific region of the workspace. Fur-
ning and subsequently acting.
thermore we show how the rotation of the input links, which
mainly actuates the yaw and pitch angles of the mechanism, also The hand-forearm assembly of the iCub humanoid robot, de-
causes unwanted coupled rotations along the roll axis. veloped at our research facility [1], has twelve independent de-
grees of freedom (DOF), weighs 0.95 kg and has a volume of
approximately 290mm x 70mm x 40mm [2]. These character-
NOMENCLATURE istics allow a considerable amount of dexterity. This dexterity,
A,B,C Homogenous transformation matrices however, comes at the price of a limited robustness and great me-
Taxis Homogenous translation matrix chanical complexity. For this reason, in recent years our group
Raxis Homogenous rotation matrix has been working on improving the dependability of this sub-
α, γ Angular geometric parameters assembly. Our first efforts were devoted to the simplification of
the iCub hand where a cost-effective design with reduced num-
ber of components and improved workspace was proposed in [3].
∗ Address all correspondence to this author.

1 Copyright © 2018 ASME


these we can cite the wrist of the robot AILA [11], the ankle of
WABIAN-2RIII [12] and the wrist of Roboray [13]. Our prelim-
inary implementations of this class of mechanisms for the iCub
wrist showed that this approach was not actually feasible because
of three main reasons, namely: i) the large amount of volume
necessary for the linear ball-screw stages, ii) the large range of
motion limitations caused by the mechanism’s self collisions and
iii) the presence of kinematic singularities in the workspace.
We then turned to a class of fully decoupled 2DOF PKMs
that provide hemispherical workspace. Some novel designs with
kinematic chains comprising of combinations of revolute and
prismatic joints were proposed in [14, 15]. One of the most
prominent work was the OmniWrist-III [16] mechanism by Ross-
Hime Designs, Inc.; which falls under the class of N-UU mecha-
FIGURE 1: CAD MODEL OF A 2DOF GIMBAL SYSTEM. nisms. With N identical limbs comprising of two universal joints
or four revolute joints, the N-UU class of mechanisms seem most
promising so far because of their mirror-symmetric architecture,
With this work, we shift our focus to the wrist. A desired wrist large workspace, hemispherical motion capabilities, and slender
mechanism shall allow large, singularity-free rotations of at least form factor of the overall system. The current work presents a
2 DOF with simpler control strategies. comprehensive analysis of different aspects of the kinematics of
There has been significant research towards the design of N-UU class of mechanisms, namely the effect of several mecha-
robotic wrists over the years and the literature is rather large [4]. nism parameters on the overall kinematics and the skewing and
Early studies presented the use of a redundant spherical wrist anisotropy of the workspace that, to our knowledge, have not yet
with 4-R serial chain where each revolute joint intersects at a been reported in the robotics literature for this type of mecha-
common point; kinematically equivalent to a spherical joint [5, nisms [17–19].
6]. A conceptual design to achieve unbounded joint limits by In the following sections of this paper, the Omni-Wrist III
replacing the intermediate joint of an Euler angle wrist with a mechanism is introduced and a kinematic model is developed,
four-bar linkage was proposed in [7] but it’s practical implemen- a CAD-based modelling and simulation approach is described
tations showed considerable restrictions on the workspace. A and detailed workspace and isotropy analyses with different geo-
standard two-axis gimbal system, as shown in the Fig. 1, also metric parameters are presented. This article concludes with the
becomes one of the predominant choices for it’s wide range of inferences from these analyses.
decoupled yaw/pitch motions, fully isotropic workspace and a
straightforward kinematics [8, 9].
Our system (the iCub platform), however, imposes addi- KINEMATIC MODELLING
tional constraints on the volume available for the mechanism. Omni-Wrist III mechanism is a N-UU type PKM with a
Indeed, the available volume can be approximated as a truncated moving platform connected to a fixed base through three or four
cone with top and base diameters of 70mm and 50mm respec- identical limbs, each comprising of a serial chain of four non-
tively and a height of 150mm. Since it is not feasible to have coplanar revolute joints. The first two joints of each limb inter-
four actuators in a limited space or even place two of them at sect at point O, the centre of the fixed base while the last two
the joint axes as in the case of a gimbal, none of the above- intersect at point P, the centre of the moving platform. The mid-
mentioned mechanisms seems to be compatible with our design dle two joints of each limb also intersect in points Ri equidistant
requirements. Moreover, these are serial kinematic chains where to the centres of the both base and the platform as shown in the
the actuator of the second axis acts as an additional inertial load Fig. 2. The mechanism can be actuated using the first joints of
for the first one [10]. We, therefore, started considering the adop- any two adjacent limbs.
tion of parallel kinematic mechanisms (PKMs). Despite having The analytical kinematic model for this mechanism was de-
limited workspaces and complex direct kinematics, PKMs of- veloped following the approach introduced in [16]. We chose to
fer several advantages over their serial counterparts, that make maintain the same axis and joint naming convention, and coordi-
them attractive for applications in robot wrists. Among these, we nate systems orientations. We wrote the loop closure equations
can cite their high stiffness, high accuracy, compact and light-to- by combining a series of homogenous transformation matrices
weight design, and synergistic behaviour in load distribution. mapping the platform position and orientation with respect to
The humanoid robotics literature is rich of examples of the base as a function of the limb joint angles, θ j , ∀ j = 1, ..., 8.
2DOF gimbal-based PKMs driven by linear actuators; among The system geometry is defined by the geometric parameters α, γ

2 Copyright © 2018 ASME


where;

B5 (θ5 ) = RX5 (γ)TZ5 (l1 )RZ5 (θ5 ) ,


B6 (θ6 ) = TYZ6 (l1 , l2 )RY6 (θ6 ) ,
B7 (θ7 ) = RZ67 (α)TX7 (l3 )RZ76 (α)RY7 (θ7 ) ,
B8 (θ8 ) = TYZ8 (l1 , l2 )RZ8 (θ8 )TZ8 (l1 )RX8 (γ) .

In both Eqn. (1) and Eqn. (2) above, the matrix Taxis (l) rep-
resents homogenous translation matrix along the specified Carte-
sian axis for a length of l units. Similarly, matrix Raxis (θ ) rep-
resents the homogeneous rotation matrix about the specified axis
by θ radians.
Now, the point P can also be directly expressed in the roll-
yaw-pitch system as given by Eqn. (3),
FIGURE 2: CAD MODEL OF A 4-UU TYPE PARALLEL
MECHANISM. C = TXYZ (x, y, z)RZ (a)RY (o)RX (n) ; (3)

where the x, y, z represent the position coordinates along the re-


and l1 , l2 , l3 . The angle α is the one between the middle joints spective Cartesian axes and n, o, a are the respective roll, yaw
for each limb, that is, axis 2/axis 3 and axis 6/axis 7. The an- and pitch orientation angles. We, therefore have the following
gle γ represents the angular offset between the two limbs; in the closed-form :
hypothesis of equally spaced “limbs” this parameter also defines
the total number of limbs in the system. The lengths l1 , l2 and A=B=C (4)
l3 are translational offsets in the defined coordinate frames. The
geometry for the L-shaped link of the limb is considered with
the following ratio, l2 = 2l1 . Also, l3 can be expressed as a func- The system of equations given in Eqn. (4) was used for nu-
tion of l1 and α as l3 = 0.67l1 [sin(α/2) + tan(α/2)]. Parameters merically solving the passive joints and subsequently the forward
l1 , l2 and l3 , were scaled in order to obtain a unit distance from kinematics using symbolic computations on SymPy [20]. The
the centre of the moving platform (point P) to the centre of the corresponding transformation matrices mentioned in Eqn. (1),
mechanism base (point O)1 . Eqn. (2) and Eqn. (3) are provided in the Appendix A.
The corresponding transformations through Limb A and
Limb B are represented by Eqn. (1) and Eqn. (2) respectively.
CAD MODELLING & SIMULATION
Obtaining closed-form analytical solutions for the kinemat-
A = A1 (θ1 )A2 (θ2 )A3 (θ3 )A4 (θ4 ) (1) ics of PKM, especially in case of N-UU, is usually a rather intri-
cate, error-prone process. To facilitate the analysis of the mech-
where; anism and the verification of expressions derived previously, we
adopted a computer-aided design (CAD) approach. We created
CAD model of the mechanism using PTC Creo Parameteric 3.0
A1 (θ1 ) = TZ1 (l1 )RZ1 (θ1 ) ,
and more specifically the Creo Mechanism multibody simulation
A2 (θ2 ) = TYZ2 (l1 , l2 )RY2 (θ2 ) , environment; the model is as shown in the Fig. 2. This model
A3 (θ3 ) = RZ23 (α)TX3 (l3 )RZ32 (α)RY3 (θ3 ) , was then simulated using the input data scanning all the combi-
A4 (θ4 ) = TYZ4 (l1 , l2 )RZ4 (θ4 )TZ4 (l1 ) . nations along the entire allowable range of the actuators (that is
±45°2 for both θ1 , θ5 ). The platform Cartesian coordinates and
the orientation angles were then recorded for each input condi-
tion. Results obtained from the CAD simulations were then com-
B = B5 (θ5 )B6 (θ6 )B7 (θ7 )B8 (θ8 ) (2) pared with those obtained from the analytical computations. As

1 Given α, γ and a unit OP distance the values of l , l and l and their ratios 2 Throughout this article we refer to angular dimensions in ° instead of [rad],
1 2 3
are univocally determined. as would be required with the SI metric system for better clarity and intuitiveness.

3 Copyright © 2018 ASME


Normalized Cartesian_Workspace

Z- Axis
Normalized Euclidean Distance Error 1.00 0.75 0.50 0.25 0.00 0.25 0.50 0.75 1.00 Alpha45_Magnitude
0.00048 1.00 24.74
40 40
0.00042
30 0.75
30 19.80
20 0.00036 0.50 20
10 0.00030 0.25 10 14.85

[°]
[°]

0 0.00024 0

Y- Axis

5
5

0.00
10 0.00018 10 9.90
0.25
20 0.00012 20
30 0.50
30 4.95
0.00006
40 0.75 40
0.00000 0.00
40 20 0 20 40 1.00
40 20 0 20 40
1 [°] 1 [°]
0.6
0.7
0.8
0.9
1.0
1.1
1.2

is
1.3

Ax
X-
(a) (c)
(b)

FIGURE 3: (a) NORMALIZED EUCLIDEAN DISTANCE ERROR BETWEEN CAD SIMULATION & ANALYTICAL COM-
PUTATION. (b) 3D VIEW OF THE CARTESIAN WORKSPACE FOR A 4-UU MECHANISM (α =45°). AN ASYMMET-
RIC BEHAVIOUR CAN BE OBSERVED AS THE SCATTER POINTS DIVERGE IN THE BOTTOM LEFT CORNER OF THE
WORKSPACE. (c) DISTANCE OF THE CENTRE OF TOP PLATFORM FROM THE BASE (MAGNITUDE).

shown in the Fig. 3a, results from both methods nicely overlap Finally to characterize the system’s properties it is also im-
with euclidean distance error in point P (the centre of top plat- portant to study and compare the dexterity measures for all of the
form) of the order of 5e−4 or less for all configurations analyzed. above cases. We computed the mechanisms’ isotropy (as defined
Indeed the CAD-based method proved to be extremely conve- in [21]) which is given by:
nient for rapidly assessing the properties of the mechanism, and
will be our default choice for future analyses. Also, the CAD- M
p
m
det(JJT )
based analysis can be very helpful in visualizing and detecting ∆= = (5)
Ψ trace(JJT )/m
possible collisions and thus fast-tracking the design process.

where J is the mechanism’s Jacobian (restricted to the yaw and


WORKSPACE & ISOTROPY ANALYSES pitch components) and m is the order of the task space. The Jaco-
The Cartesian workspace, as seen from the top, for a 4-UU bians were computed numerically from the values shown in the
mechanism obtained from one of the CAD simulations is shown tables above with the NumPy gradient function that implements
in Fig. 3b. the second order accurate central differences method [22]. The
To analyze this asymmetric behaviour in detail and to study isotropy plots for the respective cases are shown in the Fig. 4 and
the effect of geometric parameters, we considered different cases Fig. 5.
for the N-UU mechanism. Firstly, keeping α constant, we con-
sider γ=120°and γ=90°, which also corresponds to 3-UU and 4-
CONCLUSIONS
UU mechanisms respectively. Our design requirements, as stated
Several observations can be drawn from the results presented
before, are similar to a standard 2 DOF gimbal system; hence,
above:
it is essential here to also analyze the workspace of the gim-
bal mechanism. These results are presented as contours plot- 1. The plots for all values of N-UU mechanisms are not
ted against the input joint angles in a tabular form in Tab. 1 and symmetric with respect with the zero-abscissa and zero-
Tab. 2. Each row represents the contours for the normalized X, ordinate axes; we informally call this effect “warping” of
Y, Z coordinates and the Roll, Yaw, Pitch angles in the respective the workspace; interestingly the warping amplifies when α
tables for a standard Gimbal system, a 3-UU and a 4-UU mech- increases;
anism corresponding to each column in the tables. A similar set 2. Differently from the case of the gimbal, the plots of the pitch
of contour plots, as presented in Tab. 3 and Tab. 4, correspond and yaw angles of N-UU mechanisms show a dependence to
to the cases of 4-UU mechanism where the effect of α is ana- the input angles θ1 and θ5 ; we can thus confirm that the pitch
lyzed for different values, namely 30°, 45°and 60°, keeping γ and yaw DOF are not fully decoupled;
constant (=90°). As we focus on the overall behaviour of the 3. The plots of the pitch and yaw angles of N-UU mechanisms
mechanism throughout the workspace, the inter-limb collisions show that there exists a situation where the motion of the
were not taken into account for these simulations. yaw axis is independent of the input variable θ5 and where

4 Copyright © 2018 ASME


TABLE 1: CONTOUR PLOTS COMPARING THE NORMALIZED PLATFORM COORDINATES OF A 3-UU AND A 4-UU MECH-
ANISM WITH THOSE OF A STANDARD 2DOF GIMBAL SYSTEM.

Gimbal γ = 120° (3-UU) γ = 90° (4-UU)

2DOF_Gimbal_X-Coordinate Gamma120_Platform_Coordinate-X Gamma90_Platform_Coordinate-X


1.0 1.0 1.0
40 20 40
30 0.9 15 0.9 30 0.9
X-Coordinate

20 0.8 10 0.8 20 0.8


10 5 10
0.7 0.7 0.7
[°]

[°]

[°]
0 0 0
2

5
0.6 0.6 0.6
10 5 10
20 0.5 10 0.5 20 0.5
30 0.4 15 0.4 30 0.4
40 20 40
0.3 0.3 0.3
40 20 0 20 40 20 10 0 10 20 40 20 0 20 40
1 [°] 1 [°] 1 [°]

2DOF_Gimbal_Y-Coordinate Gamma120_Platform_Coordinate-Y Gamma90_Platform_Coordinate-Y


40 0.6 20 0.6 40 0.6
30 15 30
Y-Coordinate

0.4 0.4 0.4


20 10 20
10 0.2 5 0.2 10 0.2
[°]

[°]

[°]
0 0.0 0 0.0 0 0.0
2

5
10 0.2 5 0.2 10 0.2
20 10 20
0.4 0.4 0.4
30 15 30
40 0.6 20 0.6 40 0.6
40 20 0 20 40 20 10 0 10 20 40 20 0 20 40
1 [°] 1 [°] 1 [°]

2DOF_Gimbal_Z-Coordinate Gamma120_Platform_Coordinate-Z Gamma90_Platform_Coordinate-Z


40 0.6 20 0.6 40 0.6
30 15 30
Z-Coordinate

0.4 0.4 0.4


20 10 20
10 0.2 5 0.2 10 0.2
[°]

[°]

[°]
0 0.0 0 0.0 0 0.0
2

10 0.2 5 0.2 10 0.2


20 10 20
0.4 0.4 0.4
30 15 30
40 0.6 20 0.6 40 0.6
40 20 0 20 40 20 10 0 10 20 40 20 0 20 40
1 [°] 1 [°] 1 [°]

the motion of the pitch axis is independent of the input vari- Furthermore, simulations confirmed that the distance of the
able θ1 ; once again this value increases with α; centre of the top-platform from the base centre remains constant,
4. Differently from the case of the gimbal, the plots of the roll as shown in the Fig. 3c.
angles for N-UU mechanisms are not zero and reach signif-
The non-linear behaviours described above at points 1,2 and
icant values at the extremes of the mechanism’s workspace;
6 have significant consequences for the actual mechanism imple-
this means that pitch and yaw rotations are accompanied by
mentation and control. Firstly, the same control input given to the
rotations on the roll axis that amplify as the distance from
system in two different configurations will yield significantly dif-
the zero-configuration increases;
ferent output motions. This issue could, in theory, be solved by
5. The non-linear mechanism behaviour are not centered
using configuration-dependent actuator PID gains, but this would
and not symmetric with respect to the mechanism zero-
imply a substantial complication of our current robot control in-
configuration ([θ1 , θ5 ] = [0, 0]);
frastructure. Secondly, the additional rotation along the roll axis
6. Anisotropy is observed in the mechanism, especially in the
(point 4), is undesirable in our application, and we would need to
extremes of the workspace; and in accordance with the
suppress it by constructing a decoupling mechanism that would
“warping” effect, the anisotropy also amplifies with the in-
further complicate the mechanical implementation of the system.
crease in α.
For these reasons, and given our current design constraints,

5 Copyright © 2018 ASME


TABLE 2: CONTOUR PLOTS COMPARING THE PLATFORM ORIENTATION ANGLES OF A 3-UU AND A 4-UU MECHANISM
WITH THOSE OF A STANDARD 2DOF GIMBAL SYSTEM.

Gimbal γ = 120° (3-UU) γ = 90° (4-UU)

2DOF_Gimbal_Roll Gamma120_Platform_Orientation-Roll Gamma90_Platform_Orientation-Roll


90 90 90
40 20 40
30 60 15 60 30 60
20 10 20
30 30 30
Roll [°]

10 5 10
[°]

[°]

[°]
0 0 0 0 0 0
2

5
10 5 10
30 30 30
20 10 20
30 60 15 60 30 60
40 20 40
90 90 90
40 20 0 20 40 20 10 0 10 20 40 20 0 20 40
1 [°] 1[°] 1 [°]

2DOF_Gimbal_Yaw Gamma120_Platform_Orientation-Yaw Gamma90_Platform_Orientation-Yaw


90 90 90
40 20 40
30 60 15 60 30 60
20 10 20
30 30 30
Yaw [°]

10 5 10
[°]

[°]

[°]
0 0 0 0 0 0
2

5
10 5 10
30 30 30
20 10 20
30 60 15 60 30 60
40 20 40
90 90 90
40 20 0 20 40 20 10 0 10 20 40 20 0 20 40
1 [°] 1[°] 1 [°]

2DOF_Gimbal_Pitch Gamma120_Platform_Orientation-Pitch Gamma90_Platform_Orientation-Pitch


120 120 120
40 20 40
30 80 15 80 30 80
20 10 20
40 40 40
Pitch [°]

10 5 10
[°]

[°]

[°]
0 0 0 0 0 0
2

10 5 10
40 40 40
20 10 20
30 80 15 80 30 80
40 20 40
120 120 120
40 20 0 20 40 20 10 0 10 20 40 20 0 20 40
1 [°] 1[°] 1 [°]

we consider N-UU type mechanisms not suitable for our appli- (Humanoids), 2010 10th IEEE-RAS International Confer-
cation. Further research efforts will be devoted to achieving high ence on, IEEE, pp. 186–191.
dexterity wrists with a large range of motion possibly not affected [3] Sureshbabu, A. V., Metta, G., and Parmiggiani, A., 2015.
by “asymmetric” and “warped” workspaces. “A new cost effective robot hand for the iCub humanoid”.
In Humanoid Robots (Humanoids), 2015 IEEE-RAS 15th
International Conference on, IEEE, pp. 750–757.
[4] Rosheim, M. E., 1994. Robot evolution: the development
REFERENCES of anthrobotics. John Wiley & Sons, Aug.
[1] Parmiggiani, A., Maggiali, M., Natale, L., Nori, F., [5] Long, G. L., Paul, R. P., and Fisher, W. D., 1989. “The
Schmitz, A., Tsagarakis, N., Victor, J. S., Becchi, F., San- Hamilton wrist: a four-revolute-joint spherical wrist with-
dini, G., and Metta, G., 2012. “The design of the iCub hu- out singularities”. In Robotics and Automation, 1989. Pro-
manoid robot”. International journal of humanoid robotics, ceedings., 1989 IEEE International Conference on, IEEE,
9(04), p. 1250027. pp. 902–907.
[2] Schmitz, A., Pattacini, U., Nori, F., Natale, L., Metta, G., [6] Farhang, K., and Zargar, Y., 1999. “Design of spherical
and Sandini, G., 2010. “Design, realization and sensoriza- 4R mechanisms: function generation for the entire mo-
tion of the dexterous iCub hand”. In Humanoid Robots

6 Copyright © 2018 ASME


TABLE 3: CONTOUR PLOTS COMPARING THE NORMALIZED PLATFORM COORDINATES OF A 4-UU MECHANISM FOR
DIFFERENT VALUES OF THE PARAMETER α.

α = 30° α = 45° α = 60°

Alpha30_Platform_Coordinate-X Alpha45_Platform_Coordinate-X Alpha60_Platform_Coordinate-X


1.0 1.0 1.0
40 40 40
30 0.9 30 0.9 30 0.9
X-Coordinate

20 0.8 20 0.8 20 0.8


10 10 10
0.7 0.7 0.7
[°]

[°]

[°]
0 0 0
5

5
0.6 0.6 0.6
10 10 10
20 0.5 20 0.5 20 0.5
30 0.4 30 0.4 30 0.4
40 40 40
0.3 0.3 0.3
40 20 0 20 40 40 20 0 20 40 40 20 0 20 40
1 [°] 1 [°] 1 [°]

Alpha30_Platform_Coordinate-Y Alpha45_Platform_Coordinate-Y Alpha60_Platform_Coordinate-Y


40 0.6 40 0.6 40 0.6
30 30 30
Y-Coordinate

0.4 0.4 0.4


20 20 20
10 0.2 10 0.2 10 0.2
[°]

[°]

[°]
0 0.0 0 0.0 0 0.0
5

5
10 0.2 10 0.2 10 0.2
20 20 20
0.4 0.4 0.4
30 30 30
40 0.6 40 0.6 40 0.6
40 20 0 20 40 40 20 0 20 40 40 20 0 20 40
1 [°] 1 [°] 1 [°]

Alpha30_Platform_Coordinate-Z Alpha45_Platform_Coordinate-Z Alpha60_Platform_Coordinate-Z


40 0.6 40 0.6 40 0.6
30 30 30
Z-Coordinate

0.4 0.4 0.4


20 20 20
10 0.2 10 0.2 10 0.2
[°]

[°]

[°]
0 0.0 0 0.0 0 0.0
5

10 0.2 10 0.2 10 0.2


20 20 20
0.4 0.4 0.4
30 30 30
40 0.6 40 0.6 40 0.6
40 20 0 20 40 40 20 0 20 40 40 20 0 20 40
1 [°] 1 [°] 1 [°]

tion cycle”. Journal of Mechanical Design, 121(4), Dec., and Robotics, 8(2), Feb., p. 021027.
pp. 521–528. [11] Lemburg, J., de Gea Fernández, J., Eich, M., Mronga, D.,
[7] Yang, D. C., and Rauchfuss, J. W., 2001. “A new zero- Kampmann, P., Vogt, A., Aggarwal, A., Shi, Y., and Kirch-
dimension robot wrist: Design and accessibility analysis”. ner, F., 2011. “AILA-Design of an autonomous mobile
The International Journal of Robotics Research, 20(2), dual-arm robot”. In Proc. IEEE Int. Conf. on Robotics and
Feb., pp. 163–173. Automation (ICRA), IEEE, pp. 5147–5153.
[8] Stanisic, M. M., and Duta, O., 1990. “Symmetrically actu- [12] Otani, T., Iizuka, A., Takamoto, D., Motohashi, H., Kishi,
ated double pointing systems: The basis of singularity-free T., Kryczka, P., Endo, N., Jamone, L., Hashimoto, K.,
robot wrists”. IEEE Transactions on Robotics and Automa- Takashima, T., et al., 2013. “New shank mechanism for
tion, 6(5), Oct, pp. 562–569. humanoid robot mimicking human-like walking in hori-
[9] Duta, O., and Stanisic, M. M., 1989. Dextrous spherical zontal and frontal plane”. In Robotics and Automation
robot wrist, Nov. US Patent 4,878,393. (ICRA), 2013 IEEE International Conference on, IEEE,
[10] Yu, J., Wu, K., Zong, G., and Kong, X., 2016. “A compar- IEEE, pp. 667–672.
ative study on motion characteristics of three two-degree- [13] Kim, Y. J., Lee, Y., Kim, J., Lee, J. W., Park, K. M.,
of-freedom pointing mechanisms”. Journal of Mechanisms Roh, K. S., and Choi, J. Y., 2014. “RoboRay hand: A

7 Copyright © 2018 ASME


TABLE 4: CONTOUR PLOTS COMPARING THE PLATFORM ORIENTATION ANGLES OF A 4-UU MECHANISM FOR DIF-
FERENT VALUES OF THE PARAMETER α.

α = 30° α = 45° α = 60°

Alpha30_Platform_Orientation-Roll Alpha45_Platform_Orientation-Roll Alpha60_Platform_Orientation-Roll


90 90 90
40 40 40
30 60 30 60 30 60
20 20 20
30 30 30
10 10 10
Roll

[°]

[°]

[°]
0 0 0 0 0 0
5

5
10 10 10
30 30 30
20 20 20
30 60 30 60 30 60
40 40 40
90 90 90
40 20 0 20 40 40 20 0 20 40 40 20 0 20 40
1 [°] 1[°] 1 [°]

Alpha30_Platform_Orientation-Yaw Alpha45_Platform_Orientation-Yaw Alpha60_Platform_Orientation-Yaw


90 90 90
40 40 40
30 60 30 60 30 60
20 20 20
30 30 30
10 10 10
Yaw

[°]

[°]

[°]
0 0 0 0 0 0
5

5
10 10 10
30 30 30
20 20 20
30 60 30 60 30 60
40 40 40
90 90 90
40 20 0 20 40 40 20 0 20 40 40 20 0 20 40
1 [°] 1[°] 1 [°]

Alpha30_Platform_Orientation-Pitch Alpha45_Platform_Orientation-Pitch Alpha60_Platform_Orientation-Pitch


120 120 120
40 40 40
30 80 30 80 30 80
20 20 20
40 40 40
10 10 10
Pitch

[°]

[°]

[°]
0 0 0 0 0 0
5

10 10 10
40 40 40
20 20 20
30 80 30 80 30 80
40 40 40
120 120 120
40 20 0 20 40 40 20 0 20 40 40 20 0 20 40
1 [°] 1[°] 1 [°]

highly backdrivable robotic hand with sensorless contact vices. Part I: Laser beam steering devices- mathematical
force measurements”. In 2014 IEEE International Confer- modeling”. IEEE Transactions on Aerospace and Elec-
ence on Robotics and Automation (ICRA), pp. 6712–6718. tronic Systems, 42(2), Apr., pp. 718–725.
[14] Carricato, M., and Parenti-Castelli, V., 2004. “A novel fully [17] Dong, X., Yu, J., Chen, B., and Zong, G., 2012. “Geo-
decoupled two-degrees-of-freedom parallel wrist”. The In- metric approach for kinematic analysis of a class of 2-DOF
ternational Journal of Robotics Research, 23(6), June, rotational parallel manipulators”. Chinese Journal of Me-
pp. 661–667. chanical Engineering, 25(2), Mar, pp. 241–247.
[15] Li, W., He, K., Qu, Y., Zhang, J., and Du, R., 2007. [18] Yu, J., Dong, X., Pei, X., and Kong, X., 2012. “Mobility
“HEMISPHERE, a fully decoupled parallel 2-DOF spher- and singularity analysis of a class of two degrees of free-
ical mechanism”. In Proc 7th WSEAS Int Conf Robotics, dom rotational parallel mechanisms using a visual graphic
Control & Manufacturing Technology, Stevens Point, RO- approach”. Journal of Mechanisms and Robotics, 4(4),
COM’07, World Scientific and Engineering Academy and Sept., p. 041006.
Society (WSEAS), pp. 301–306. [19] Wu, Y., and Carricato, M., 2017. “Synthesis and singular-
[16] Sofka, J., Skormin, V., Nikulin, V., and Nicholson, D., ity analysis of N-UU parallel wrists: A symmetric space
2006. “Omni-Wrist III- a new generation of pointing de- approach”. Journal of Mechanisms and Robotics, 9(5),

8 Copyright © 2018 ASME


2DOF_Gimbal_Isotropy Gamma120_Isotropy Gamma90_Isotropy
1.0 1.0 1.0
40 20 40
30 0.8 15 0.8 30 0.8
20 10 20
10 0.6 5 0.6 10 0.6
[°]

[°]

[°]
0 0 0
2

5
10 0.4 5 0.4 10 0.4
20 10 20
30 0.2 15 0.2 30 0.2
40 20 40
0.0 0.0 0.0
40 20 0 20 40 20 10 0 10 20 40 20 0 20 40
1[°] 1[°] 1[°]

(a) Gimbal (b) γ = 120° (3-UU) (c) γ = 90° (4-UU)

FIGURE 4: MECHANISM ISOTROPY FOR GIMBAL SYSTEM AND N-UU MECHANISMS WITH DIFFERENT VALUES OF γ.

Alpha30_Isotropy Alpha45_Isotropy Alpha60_Isotropy


1.0 1.0 1.0
40 40 40
30 0.8 30 0.8 30 0.8
20 20 20
10 0.6 10 0.6 10 0.6
[°]

[°]

[°]
0 0 0
5

5
10 0.4 10 0.4 10 0.4
20 20 20
30 0.2 30 0.2 30 0.2
40 40 40
0.0 0.0 0.0
40 20 0 20 40 40 20 0 20 40 40 20 0 20 40
1[°] 1[°] 1[°]

(a) α = 30° (b) α = 45° (c) α = 60°

FIGURE 5: MECHANISM ISOTROPY FOR A 4-UU MECHANISM WITH DIFFERENT VALUES OF α.

Aug., p. 051013. Appendix A: Transformations for the Kinematic Model


[20] Meurer, A., Smith, C. P., Paprocki, M., ertk, O., Kirpichev, The individual translation and rotation matrices3 for Eqn. (1)
S. B., Rocklin, M., Kumar, A., Ivanov, S., Moore, J. K., and Eqn. (2) of the kinematic model, are given as follows:
Singh, S., Rathnayake, T., Vig, S., Granger, B. E., Muller,
R. P., Bonazzi, F., Gupta, H., Vats, S., Johansson, F., Pe-  
dregosa, F., Curry, M. J., Terrel, A. R., Rouka, t., Saboo, 1 0 00
0 1 0 0
A., Fernando, I., Kulal, S., Cimrman, R., and Scopatz, A., TZ1 = TZ4 = TZ5 = TZ8 =   (6)
0 0 1 0
2017. “SymPy: symbolic computing in python”. PeerJ
Computer Science, 3, Jan., p. e103. 0 0 0 l1
[21] Kim, J.-O., and Khosla, K., 1991. “Dexterity measures for
design and control of manipulators”. In Intelligent Robots
and Systems ’91. ’Intelligence for Mechanical Systems,
Proceedings IROS ’91. IEEE/RSJ International Workshop 
C(θ1 − α2 ) −S(θ1 − α2 ) 0 0

on, pp. 758–763 vol.2.  S(θ1 − α ) C(θ1 − α ) 0 0
RZ1 =  2 2 (7)
[22] van der Walt, S., Colbert, S. C., and Varoquaux, G., 2011.  0 0

1 0
“The NumPy array: A structure for efficient numerical 0 0 01
computation”. Computing in Science Engineering, 13(2),
March, pp. 22–30.

3 To simplify the notations here, the sine() and cosine() functions are abbrevi-

ated to S() and C() respectively.

9 Copyright © 2018 ASME


 
C(θ5 − α2 ) −S(θ5 − α2 ) 0 0
 
C(θ8 − α2 ) −S(θ8 − α2 ) 0 0
 S(θ5 − α ) C(θ5 − α ) 0 0  S(θ8 − α ) C(θ8 − α ) 0 0
RZ5 =  2 2 (8) 2 2
0 0

1 0 RZ8 =   (16)
  0 0 1 0
0 0 01 0 0 01

 
100 0    
0 1 0 l2  1 0 0 0 1 0 0 0
TYZ2 = TYZ6 =   (9) 0 Cγ −Sγ 0 0 Cγ Sγ 0
0 0 1 −l1  RX5 =  ;R = (17)
0 Sγ Cγ 0 X8 0 −Sγ
   
Cγ 0
000 1
0 0 0 1 0 0 0 1

    Similarly, the matrices corresponding to Eqn. (3) can be


Cθ2 0 Sθ2 0 Cθ6 0 Sθ6 0 written as follows:
 0 1 0 0  0 1 0 0
RY2 =  ; R =  (10)
0 Y6 −Sθ6

−Sθ2 0 Cθ2 0 Cθ6 0    
1 0 0 x Ca −Sa 0 0
0 0 0 1 0 0 0 1 0 1 0 y ; RZ =  Sa Ca 0 0
 
TXYZ = 
0 (18)
0 1 z   0 0 1 0
0 0 0 1 0 0 01
 α 
C( 2 ) −S( α2 ) 0 0
 S( α ) C( α ) 0 0
RZ23 = RZ32 = RZ67 = RZ76 =  2 2  (11)
 0 0 1 0    
0 0 01 Co 0 So 0 1 0 0 0
 0 1 0 0 0 Cn −Sn 0
−So 0 Co 0 ; RX = 0 Sn Cn 0
RY =     (19)
  0 0 0 1 0 0 0 1
1 0 0 l3
0 1 0 0
TX3 = TX7 = 
0
 (12) To compute the forward pose kinematics for a given input
0 1 0
anlges θ1 , θ5 , we need to solve for the other passive angles. For
0 0 01
θ2 , θ6 , we extract the elements from the last column of the matri-
ces A and B in the loop closure Eqn. (4) and solve it numerically
on SymPy. Owing to the mechanism symmetry, the following
   
Cθ3 0 Sθ3 0 Cθ7 0 Sθ7 0 relations between the other joint angles can also be considered
 0 1 0 0  0 1 0 0 (see [19] for details):
RY3 =  ;R =  (13)
0 Y7 −Sθ7

−Sθ3 0 Cθ3 0 Cθ7 0
0 0 0 1 0 0 0 1 θ3 = θ2 ; θ7 = θ6 ; θ4 = θ1 ; θ8 = θ5 . (20)

 
100 0
0 1 0 −l2   
TYZ4 = TYZ8 =   (14) CaCo CaSoSn − SaCn CaSoCn + SaSn x
0 0 1 −l1   SaCo SaSoSn +CaCn SaSoCn −CaSn y
000 1 C=
 −So
 (21)
CoSn CoCn z
0 0 0 1

 
C(θ4 − α2 ) −S(θ4 − α2 ) 0 0 Finally, to obtain the platform coordinates and it’s roll, yaw,
 S(θ4 − α ) C(θ4 − α ) 0 0 pitch orientations, we equate the elements of matrices A and B
RZ4 =  2 2  (15)
 0 0 1 0 with those of the matrix C as given by Eqn. (21) and solve them
0 0 01 numerically.

10 Copyright © 2018 ASME

View publication stats

Das könnte Ihnen auch gefallen