IEEE Robotics & Automation Magazine - March 2017 - 33

amount of time, effort, and money to fabricate custom
hands or purchasing from a limited and expensive set of
commercially available options [3]-[9]. Researchers may not
have the resources or ability for the former, and in the latter
case, they would be both unable to modify their designs and
also dependent on the manufacturer for repairs and maintenance. Among other challenges, this situation prevents hardware and software research in manipulation from co-evolving.
The increased accessibility of 3-D printing, in particular
FDM thanks in part to open-source initiatives [10], has made it
more tractable to expediently produce custom parts on
demand [11]. However, these parts are generally nonmoving,
nonload-bearing, and made of a single homogeneous material.
To provide the community with the ability to 3-D print inexpensive, customizable, and easy-to-fabricate components that
are load-bearing and articulated, modifications must be made
to extend existing additive manufacturing processes. As a
means to address some of those concerns, we had previously
developed a fabrication technique called HDM, in which
3-D-printed parts are combined with additional, deposited resins to produce monolithic, multimaterials parts, integrating, for
example, rubber flexure joints and soft fingerpads as well as
components such as tactile sensors that are molded into the
parts. A robotic finger produced using the HDM process is
lightweight and robust and has a low part count, and its fabrication requires less than one hour of manual assembly.
Using the design guidelines established during the refinement of the HDM process, we have developed a library of
extensible, open-source hand designs that are modifications
of some of their previous underactuated hands [12] to enable
effective dissemination via additive manufacturing techniques
(Figure 1). These tendon-driven designs require only
3-D-printed components and readily available off-the-shelf
parts. The designs utilize self-contained hobby servos for
actuators, and the implementation of adaptive, underactuated
mechanisms enables a high degree of capability with only
open-loop control. The fingers utilize cast, flexural joints for
increased robustness, and they can be easily swapped out due
to their monolithic and modular design. The fabrication and
assembly processes for each of the hands are extensively documented [13] with step-by-step instruction guides and videos
to promote design improvements and adoption in various
applications by end users in a variety of research domains.
Related Work
Functional Components via 3-D Printing
Parts created via FDM have primarily been either static fixtures or simple, nonload-bearing mechanisms, but the structural integrity of FDM parts can be enhanced through various
postprocessing methods. Printed parts can be used as the limited-use mold components directly or the positive to create a
more durable mold for injection molding. Fill-compositing
[14] deposits epoxy or other resins within voids of printed
parts and can improve the overall part strength by up to 45%.
Articulated mechanisms can utilize cast flexures in place of

Figure 1. The underactuated, four-finger Model T hand mounted
on a whole-arm manipulator (WAM). This was the initial hand
design in the Yale OpenHand Project design library.

revolute joints for increased durability [15] and compliance to
minimize damage during collisions.
In robotics, relevant work has chiefly used FDM for body
frame subcomponents in direct-drive systems such as miniature humanoids, modular robotics, and legged robots. Many
of these systems are driven by hobby-grade servos and have
little to no additional transmission elements. Researchers have
also produced several proof-of-concept hand designs [16]-
[18] using FDM and related low-cost, rapid-prototyping
methods. However, to our knowledge, none have evaluated
their designs' potential for long-term use or the structural
limitations of the printed subcomponents. We will detail how
the Yale OpenHand Project builds on these past initiatives in
the section "Hand Design."
Underactuated Hand Design
The OpenHand Project implements many of the mechanical
design strategies present in the shape deposition manufacturing
(SDM) hand [12], an underactuated, tendon-driven, four-finger
hand driven by a single actuator. Underactuated adaptability, via
differential mechanisms within each finger as well as between
the fingers, enable the hand to passively conform to various
object geometries using only open-loop control. This hand was
named for its fingers' fabrication process, SDM, where alternating material removal and deposition processes are combined to
create multimaterial structures. The initial effort on the Yale
OpenHand Project sought to recreate a more compact and
simplified version of the SDM hand using 3-D printing and offthe-shelf components [19], as shown in Figure 1.
Underactuated fingers or transmissions can be found in several commercial hands [3]-[5]. Each two-link finger in the SDM
hand and Yale OpenHand iterations is driven by a single tendon,
and the final torque at each joint is determined by a combination of the actuating tendon force ^ fah, the effective pulley radii
March 2017

IEEE ROBOTICS & AUTOMATION MAGAZINE

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