Professional Documents
Culture Documents
1
Corresponding author.
Manuscript received May 29, 2008; final manuscript received January 19, 2009; Fig. 1 Closing sequence of a two-phalanx underactuated fin-
published online March 17, 2009. Review conducted by Vijay Goel. ger †5‡
Journal of Medical Devices Copyright © 2009 by ASME MARCH 2009, Vol. 3 / 011007-1
increases, the complexity of the analysis grows exponentially and a worst case value of 100 m is reported in literature 关18兴. In the
the usable workspace defined by having all-positive contact forces latter case, the removal of this recast layer should be taken into
might decrease drastically. The fingers of this gripper have five account and the joints should be expanded accordingly. Using
phalanges and thus may have a very small workspace if improp- careful very fine polishing, the lifetime of the remaining test joints
erly designed. Hence, this paper presents several new results con- reached an average of 713 cycles, a significant improvement since
cerning compliant underactuated grasping: improved joint design, this material removal could not be controlled with a precision of
novel kinematic architectures of five-phalanx fingers, and the per- half a thousandth of 1 in. 共which might explain the discrepancy
formance sensitivity analysis of the latter. with the theoretical value兲. This polishing should also be done as
much as possible in the direction of the sheet flow since sanding
2 Joint Design perpendicular to the edges of the sheet 共as well as ultrasonic
The gripper design starts with the robotic fingers. Their length cleaning兲 is reported to propagate microcracks. It should be noted
was chosen to be close to the one from some common laparo- that with the asymmetrical joint, the mechanical limit was not
scopic pincers 关12兴. Nitinol was chosen to constitute the gripper present in the test joints since it prevents polishing.
because of its biocompatibility and superelastic properties, allow-
ing the joints to have a large range of motion and to be used in 3 Underactuated Finger Design
surgery without additional coating. A wire cut electrical discharge Underactuated fingers can be analyzed as the connection of
machining 共EDM兲 process can be used to manufacture NiTi parts robotic fingers plus several differential mechanisms driving their
but it imposes a minimum joint thickness. In this case, to avoid joints, which results in a self-adaptive capability 关5兴. A finger with
performance issues such as deformation of links outside of the five phalanges was chosen in order to have a large adaptability
joints or a reduced lifetime because of joints being too thin, the and to reduce the rotation demanded from each individual joint,
minimal joint thickness is set to 0.25 mm. The shape of the hinge thereby increasing the lifetime of the device. The length of each
joints is inspired by Ref. 关13兴. A corner-filleted flexure hinge has phalanx was decided to be 5 mm and of each joint to be 1.4 mm
very good compliance for single-axis use, both in the symmetric long. These considerations bring the length of the finger 共32 mm兲
and nonsymmetric cases 关14兴. Both types of joint were modeled in line with some common laparoscopic pincers. Regarding the
and optimized using a design of experiments method 关15兴 with a driving mechanism, it is composed of a transmission mechanism
finite element method 共FEM兲 commercial software, where infinite distributing the actuation torque to the phalanges, and a base
life was considered at approximately 103 flexures as reported in mechanism converting a linear force into the input torques for the
the low-cycle fatigue data in Ref. 关16兴. The desired maximum fingers 共cf. Fig. 3兲. The objective of the transmission mechanism
rotation of the joints is 70 deg. By using a filleted flexure hinge, is to drive the finger as it grasps an object. It must continue to
the life of the joint is increased since there are less stress concen-
tration points. The results show that the maximal lifetime for the
asymmetrical joint is achieved with a fillet-radius of 0.30 mm, and
for the symmetrical joint with a fillet-radius of 0.4 mm. The re-
sulting joints are illustrated in Fig. 2.
Asymmetrical joints have been considered to avoid a foreign
object getting lodged in the joints between the phalanges contact-
ing the object seized, and thus, possibly damaging biological tis-
sue. Mechanical limits are included in these joints to prevent them
from rotating in an undesired direction. For the symmetrical joint,
the fillet-radius of 0.30 mm was chosen because it produces an
acceptable lifetime for the device, and by increasing the compli-
ance of the joint, it reduces the force needed to drive the gripper.
However, these theoretical results have to be taken with a grain
of salt. Indeed, it has been reported that the EDM machining of
NiTi alloys lead to a recast layer and microcracks, which dramati-
cally reduce life in fatigue 关17兴. Experiments have been done with
test joints in nitinol using the geometry illustrated in Fig. 2, and
indeed, the actual life in fatigue of these joints ranged between a
few dozens and 140 cycles. However, this can be improved by an
appropriate post-treatment of the joints such as polishing or
chemical etching 关18,19兴. According to the engineers from the
company that has provided us with the NiTi sheet, this recast layer
is approximately half a thousandth of 1 in. 共13 m兲 thick, while Fig. 3 Division of the constitutive mechanisms of the gripper
pared.
The use of compliant joints demands the rotation of the joints to 4 Optimization of the Transmission Mechanism
be as even as possible, distributing the deformations to as many
joints as possible to increase the life of the device. The first archi- 4.1 Introduction. The objective of this optimization proce-
tecture considered 共Fig. 4共a兲兲 consists of a five-bar linkage 关20兴 dure is to obtain the geometric parameters that will be used in the
where the proximal phalanx is subdivided into four phalanges. transmission mechanism of the fingers. These geometric param-
Each division generates a new DOF. The second architecture 共Fig. eters define the performance of the gripper inside its workspace.
4共b兲兲 is a three-stage mechanism inspired by the work presented in Each joint has a range of motion of 45 deg, lower than the de-
Ref. 关9兴. It is a good compromise between a four-stage mechanism signed limits of the joints to further increase the lifetime of the
with a better deformation distribution, and the force required to gripper. Considering the large number of parameters to be taken
drive the finger. The third and final architecture is a modification into account, the optimization was done using a genetic algorithm
of a five-bar linkage with a four-bar linkage in the middle to 共GA兲. Basic parameters for the latter were based on DeJong and
decrease the range of motion required in the joints of the trans- Spears’ settings 关22兴. The population size and the crossover frac-
mission mechanism of the former case. tion were further refined by a deterministic study where a range of
The distribution of contact forces once an object has been con- values for each of the parameters was tested against a function
tacted is much more relevant than the movement of the finger representing a five-phalanx underactuated finger using a three-
before the grasp 关21兴 and is therefore the main concern in the stage transmission architecture. Because the genetic algorithm is a
design. A method to obtain the analytical expressions of the con- stochastic system, a Monte Carlo analysis was done with ten it-
tact forces of underactuated fingers is presented in Ref. 关5兴 and erations. The average results show a marked improvement by in-
has been extended to compliant cases in Ref. 关11兴. The transmis- creasing the population size and by reducing the crossover frac-
sion and Jacobian matrices presented in these references are valu- tion compared with the recommended values.
able tools to envision the performance of the finger. Similarly, 4.2 Grading Function. When using an index to grade the
contact forces can be obtained with a static analysis, which is performance of a transmission mechanism, the finger is optimized
preferable in the compliant case due to the complexity of the latter to perform well under a certain metric. If the optimization metric
matrices. Taking into account five contact forces, the following does not favor the desired usage, the resulting gripper will under-
linear system is obtained: perform. The first attempt at optimizing this gripper had a global
Gf = k 共1兲 approach, where the resulting mechanism would perform well
throughout the whole workspace. In reality, this method is not the
where k is a vector containing the elastic torques generated by the best suited as it does not favor the most common cases, and thus,
compliance of the joints 关11兴, matrix G is obtained by combining a localized approach was chosen by developing particular test
the equations derived from the static equilibrium of the finger, and cases or objects.
vector f contains the contact forces 共fc = 关F1 ¯ F5兴T兲 and the inter- A grading or fitness function assigns an index to each design.
nal forces of the mechanism 共fi兲, i.e., Given the geometric parameters and five angles 共1 , . . . , 5兲 that
冤 冥
define a grasp, it evaluates the performance of the finger in this
Ta particular situation. First of all, this function requires the follow-
冋册 冋册
T 1 = − k 1⌬ 1 ing conditions.
fc 0
f= , k= , t = T 2 = − k 2⌬ 2 共2兲 1. All contact forces are positive. This is required to have a
fi t stable grasp 关23,24兴.
...
2. The torque acting on the last phalanx must not be negative.
T n = − k n⌬ n 3. The projection on the last phalanx of the intersection be-
where ki is the stiffness of the compliant joint associated with i tween both lines associated with the links attached to this
关11兴, and n is the number of compliant joints in the finger. Solving last phalanx is inside the latter 关7兴.
Eq. 共2兲 allows to compute the force vector f, namely, 4. When using a test object, the resultant force applied on this
object pushes the latter toward the palm of the gripper.
兺
5
Fi
i=1
mfm = 共5兲
Fin
3. Force isotropy. The standard deviation 共兲 of the forces al-
lows us to characterize the uniformity of the grasp; a smaller
standard deviation indicates a uniform pressure of the finger.
Fig. 4 Finger architectures considered This rewards force-isotropic 关20兴 designs and is desired in
Architecture 1 2 3
Fig. 5 Examples of test objects used in the optimization Together, these elements are combined with different weights
serving as the fitness function of the GA as follows:
Mechanism grade = 10x̄T + 3x̄L + x̄W 共8兲
order to avoid large local forces on the object
The grading function defined above has many different local
共fc兲 minima. Given that the probability of not finding the global mini-
mfi = 1 − 共6兲
共fc兲 + 5 mum of this function is not negligible 关25兴, the GA was repeatedly
run to collect various local minima and then compare the three
Finally, these indices are multiplied together to obtain a grade architectures considered. Figure 6 shows a box plot of the results
for the finger with the given parameters and position of the GA runs with the three different architectures where one
can appreciate the large variation between results and the large
Grade = mdmfmmfi 共7兲 number of outlier values, which are better performing geometric
parameters and thus more desirable targets. These results suggest
4.3 Optimization Process. Each set of geometric parameters a large sensibility of the optimization process to small changes in
共candidate兲 is evaluated with the grading function defined by Eq. the geometric parameters. Statistics of the grades are shown in
共7兲. Test objects were also chosen, considering that the gripper is Table 1.
believed to grasp objects with shapes close to a circle. Examples The results were interesting, and, in particular, architecture 3
of these test cases are illustrated in Fig. 5. In the optimization 共cf. Fig. 4共c兲兲 showed potential regarding its capability to enve-
process, the radius of the circle representing an object being lope the finger around an object. However, two revolute joints
A normal probability plot of the residuals of the eight-way where ␦d has a value of 0 for the finger that lags behind if it
ANOVA is shown in Fig. 8. Although the bulk of the observations receives less torque than the other, or else a value of 1. The vol-
form a straight line, implying a normal distribution, there are a ume of the surface generated by the index defined by Eq. 共9兲 when
small number of points deviating from it. The ANOVA revealed it is calculated over the whole mechanism workspace is used as
295 statistically relevant interactions 共p ⬍ 0.05兲. The p-value is the fitness function of a GA again. This surface is illustrated in
the probability that the difference between groups during the ex- Fig. 10 for the geometric parameters of the driving mechanism
periments happened by chance 关26兴. When it is lower than the presented in Table 3, which correspond to the result of the opti-
significance level, ␣ = 0.05, the factor or interaction is considered mization.
A1 B1 C1 A2 B2 C2 A3 B3 C3 ⌿
15.7 18.4 9.1 13.4 12.9 3.4 14.7 14.2 7.5 83 deg
D B C ␣
Finally, a relationship between the length of the finger and the Fig. 13 von Mises stress of a nitinol gripper during a sym-
palm reported in Ref. 关28兴 was considered during this optimiza- metrical grip
tion. Taking the suggestions proposed along with size constraints,
a 0.545 palm/finger ratio was chosen. The finger length being 32
mm 共it includes five phalanges and five flexural joints兲, the palm
length is 17.5 mm. With the finger, driving and transmission
mechanisms determined, a final design is obtained, illustrated in
Fig. 11, and a prototype is shown in Fig. 12.
6 Finite Element Simulation “Design and Application of Compliant Mechanisms for Surgical Tools,”
ASME J. Biomech. Eng., 127共6兲, pp. 981–989.
A finite element simulation was done using a commercial FEM 关2兴 Boudreault, E., 2006, “Conception de Préhenseurs Sous-Actionnés avec Ar-
software to validate the final design obtained. The software has a ticulations Déformables,” MS thesis, Université Laval, QC, Canada.
关3兴 Boudreault, E., and Gosselin, C., 2006, “Design of Sub-Centimetre Underac-
shape memory alloy 共SMA兲 module, which allows it to give an tuated Compliant Grippers,” Proceedings of the 2006 ASME International De-
accurate simulation of how an actual prototype would behave be- sign Engineering Technical Conferences and Computers and Information in
cause it takes into consideration the superelastic properties of niti- Engineering Conference, Philadelphia, PA, Paper No. DETC2006–99415.
nol. To the best of the authors’ knowledge, it is the first time that 关4兴 Dollar, A., and Howe, R. D., 2006, “Joint Coupling Design of Underactuated
Grippers,” Proceedings of the 2006 ASME International Design Engineering
this is done with an underactuated compliant gripper. To compare Technical Conferences and Computers and Information in Engineering Con-
and highlight the usefulness of nitinol, the same simulations were ference, Philadelphia, PA, Paper No. DETC2006–99645.
performed using four other different materials: aluminum alloy, 关5兴 Birglen, L., Laliberté, T., and Gosselin, C., 2008, Underactuated Robotic
316L stainless steel, polypropylene, and ABS plastic 关29兴. Results Hands, Springer, New York.
关6兴 Hirose, S., and Umetani, Y., 1978, “The Development of Soft Gripper for the
are illustrated in Figs. 13 and 14. The results are promising and Versatile Robotic Hand,” Mech. Mach. Theory, 13, pp. 351–358.
show that the gripper can envelope and secure different objects of 关7兴 Laliberté, T., and Gosselin, C., 2001, “Underactuation in Space Robotic
various shapes. For the case shown in Fig. 13, the base of the Hands,” International Symposium on Artificial Intelligence, Robotics and Au-
gripper required a displacement of 7 mm and has a theoretical tomation in Space, Montréal, Canada.
lifetime of 982 flexures. It should be noted that the actual lifetime 关8兴 Eiselstein, L., 2005, “Review of Fatigue and Fracture Behavior in NiTi,”
Medical Device Materials III-Proceedings of the Materials and Processes for
of the gripper also depends on the object共s兲 seized since each joint Medical Devices Conference, Boston, MA, pp. 135–147.
is differently solicitated with different objects. The worst case 关9兴 Gosselin, C., and Laliberté, T., 1996, “Underactuated Mechanical Finger With
scenario corresponds to an object where one joint is consistently Return Actuation,” U.S. Patent No. 5 762 390.
fully rotated. Then, the lifetime of the gripper is equal to the 关10兴 Birglen, L., and Gosselin, C., 2004, “Kinetostatic Analysis of Underactuated
Fingers,” IEEE Trans. Rob. Autom., 20共2兲, pp. 211–221.
lifetime of this joint as discussed in Sec. 2. The required force to 关11兴 Birglen, L., 2006, “An Introduction to the Analysis of Linkage-Driven Com-
grasp an object and the maximum von Mises stress are reported in pliant Underactuated Fingers,” 2006 ASME International Design Engineering
Table 4. Technical Conferences, Philadelphia, PA, Paper No. DETC2006–99048.
关12兴 Intuitive Surgical, 2008, “Intuitive Surgical-EndoWrist Instruments,” URL:
7 Conclusions http://www.intuitivesurgical.com/products/endowrist_instruments /index.aspx.
关13兴 De Bona, F., and Munteanu, M. Gh, 2005, “Optimized Flexural Hinges for
This paper underlines the versatility offered by the use of com- Compliant Micromechanisms,” Analog Integr. Circuits Signal Process., 44,
pliant underactuated grippers using nitinol. The use of such end- pp. 163–194.
关14兴 Lobontiu, N., 2002, Compliant Mechanisms: Design of Flexure Hinges, CRC,
effectors in laparoscopic surgeries is an exciting avenue of oppor- Boca Raton, FL.
tunity where the onus of the grasping motion is shifted away from 关15兴 Montgomery, D. C., 2005, Design and Analysis of Experiments, Wiley, New
the control electronics to an intelligent mechanism. York.
The development of a gripper was presented, which substanti- 关16兴 Wilkes, K., and Liaw, P., 2000, “The fatigue behavior of shape-memory al-
loys,” Journal of the Minerals, Metals and Materials Society, 52共10兲, pp. 45–
ates the feasibility of the use of underactuated compliant grippers. 51.
Special attention was paid to the selection and optimization of the 关17兴 Lin, H., Lin, K., and Cheng, I., 2001, “The Electro-Discharge Machining Char-
transmission mechanism, with a focus on finding a set of geomet- acteristics of TiNi Shape Memory Alloys,” J. Mater. Sci., 36, pp. 399–404.
ric parameters that allows the fingers to perform satisfactorily un- 关18兴 Wu, M. H., 2001, “Fabrication of Nitinol Materials and Components,” Pro-
ceedings of the 2001 International Conference on Shape Memory and Super-
der all circumstances. The results obtained were validated using a elastic Technologies, pp. 285–292.
FEM software, which took into account the superelastic properties 关19兴 Shabalovskaya, S., Anderegg, J., Laab, F., Thiel, P., and Rondelli, G., 2003,
of the material. “Surface Conditions of Nitinol Wires, Tubing, and As-Cast Alloys. The Effect
Further validation of the simulation results should be done, as of Chemical Etching, Aging in Boiling Water, and Heat Treatment,” J. Biomed.
the results reported for the complete gripper are only based on Mater. Res., Part B: Appl. Biomater., 65B共1兲, pp. 193–203.
关20兴 Birglen, L., and Gosselin, C., 2004, “Optimal Design of 2-Phalanx Underac-
numerical simulations. Part of the continuing work should include tuated Fingers,” Proceedings of the 2004 International Conference on Intelli-
destructive testing after chemical etching to confirm the lifetime gent Manipulation and Grasping, Genova, Italy, pp. 110–116.
expectancy analysis. Indeed, due to the mechanical limits between 关21兴 de Visser, H., and Herder, J. L., 2000, “Force Directed Design of a Voluntary
the phalanx joints, polishing was not possible. Hence, only the Closing Hand Prosthesis,” J. Rehabil. Res. Dev., 37共3兲, pp. 261–271.
关22兴 DeJong, K., and Spears, W., 1990, “An Analysis of the Interacting Roles of
individual joints have been tested experimentally but not the grip- Population Size and Crossover in Genetic Algorithms,” Proceedings of the
per. The sensitivity analysis presented here relied on a statistical First Workshop in Parallel Problem Solving From Nature, Springer-Verlag,
method an often used in other fields, which was applied success- Berlin, pp. 38–47.
fully to the development of an underactuated compliant gripper. 关23兴 Birglen, L., and Gosselin, C., 2003, “On the Force Capabilities of Underactu-
ated Fingers,” Proceedings of the 2003 IEEE International Conference on Ro-
This development process has indicated that the optimization cri- botics and Automation, Taipei, Taiwan, pp. 1139–1145.
teria based on the contact forces are valid but very stringent when 关24兴 Birglen, L., and Gosselin, C., 2006, “Optimally Unstable Underactuated Grip-
the number of phalanges is large. per: Synthesis and Applications,” 2006 ASME International Design Engineer-
ing Technical Conferences, Philadelphia, PA, Paper No. DETC2006–99047.
Acknowledgment 关25兴 Rudolph, G., 1994, “Convergence Analysis of Canonical Genetic Algorithms,”
IEEE Trans. Neural Netw., 5共1兲, pp. 96–101.
The support of the Natural Sciences and Engineering Research 关26兴 Dean, A. M., and Voss, D., 1999, Design and Analysis of Experiments,
Council of Canada and the Canadian Foundation for Innovation is Springer, New York.
关27兴 Birglen, L., and Gosselin, C., 2006, “Force Analysis of Connected Differential
gratefully acknowledged. The authors wish to thank Jeff Archer Mechanisms: Applications to Grasping,” Int. J. Robot. Res., 25共10兲, pp. 1033–
and Bill Hochella from Johnson Mattey Metals for their help. 1046.
关28兴 Laliberté, T., and Gosselin, C., 1998, “Simulation and Design of Underactu-
ated Mechanical Hands,” Mech. Mach. Theory, 33共1–2兲, pp. 39–57.
References 关29兴 MatWeb, 2008, “MatWeb: Material Property Data,” URL: http://www
关1兴 Kota, S., Lu, K.-J., Kreiner, Z., Trease, B., Arenas, J., and Geiger, J., 2005, .matweb.com/.