IEEE Robotics & Automation Magazine - March 2017 - 36

Table 2. A comparison of medium-complexity hands.
Hand

Number
of Fingers

Number
of Actuators

Base
Height (mm)

Base
Width (mm)

Weight (g)

Grip Force (N)

Barrett Hand [3]

75.5

130

1,200

2G Velo [29]

Robotiq (two-finger)

140

890

30-100

Robotiq (three-finger) [4]

126

2,300

15-60

Schunk SDH Hand [9]

122

1,950

i-HY [22]

105

1,390

RightHand Reflex [5]

Lacquey P102 [7]

113-203

1,250

Lacquey A101

127

1,000

Festo MultiChoiceGripper [8]

215

148

660

OpenHand Model T

100

490

11.54 ± 1.20

OpenHand Model T42

55-80

90-105

400

9.60 ± 0.25

OpenHand Model O

100-125

752

12.33 ± 0.71

Design for 3-D Printing
For this study, all printed parts were produced on a Stratasys
Fortus 250 mc, a commercial desktop FDM printer. Circular
clearances for revolute joints or press-fit components are
printed with their axes of rotation parallel to the print
direction to avoid shearing failure. Printed part geometries
were kept as basic as possible, such that the preferred print
direction could be easily identified.
Consistent with the guidelines for injection-molding, the
printed part designs also avoid sudden changes in part thick-

60 mm

MLP-25
Force
Transducer
70 mm

Figure 5. The load cell setup used to evaluate the maximum,
sustainable grasp force for each hand. All evaluated hand designs
have multilink fingers, so rounded contours were printed and
attached to the load cell to maximize contact.

IEEE ROBOTICS & AUTOMATION MAGAZINE

March 2017

800

subbodies. Under the worst-case loading scenario, which
typically would not occur at these interfaces as they are implemented in the OpenHand designs, both urethane materials
can withstand up to 100 N of pulling force prior to fully
disengaging from the printed bodies and failing [23].

10-20

ness to avoid warping or peeling from the build platform.
This is especially impactful for nonenclosed FDM machines,
such as low-cost, do-it-yourself options [10]. Load-bearing
features are always printed with solid infill and have a minimal dimension of least 3 mm in size, while sacrificial features,
like temporary mold walls, are 0.7 mm in size.
Bearing surfaces, particularly for tendons, remain a challenge for 3-D-printed parts, and abrasions from repeated
actuation cycles can wear through printed surfaces. A 100-lb
test, 0.5-mm diameter Spectra line is used for tendon-routing
in all of the hands. Consequently, the OpenHand designs
include either steel dowel pins or small nylon pulleys to route
the tendons, as shown in Figure 3. For useful adaptability and
reconfiguration, friction along the routing path needs to
be minimized.
Performance
General Comparison
Table 2 compares the basic characteristics of the Yale
OpenHand designs with that of available commercial hands.
In terms of size and weight, the OpenHand options are comparable with commercial alternatives and can serve as a dropin replacement on compatible manipulator platforms for
basic grasping tasks. The low weight, due to the use of printed
ABS, could make these designs particularly appealing for
mobile manipulation applications.
The holding grasp force for OpenHand was measured by
an MLP-25 load cell, shown in Figure 5. Printed attachments
to the load cell were used to guarantee that all finger links
made contact during the force measurement. All hands were
run at 12 V, with the servos' maximum torque limited at 40%
of their specified stall torque, as recommended by the manufacturer. The hands are capable of a much higher grasp force
than the values listed in Table 2, but the recorded values

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - March 2017