IEEE Robotics & Automation Magazine - September 2023 - 141

Another reason for ORH projects to share editable design
files is to account for inconsistencies in fabrication processes.
Processes like 3D printing vary depending on the printers
and materials used, and developers should consciously avoid
designing hard-to-fabricate features in their custom parts. One
way to avoid such variations altogether in fabrication processes
is to rely more on commercial off-the-shelf (COTS) components.
But, even for purchased components, developers should
consider alternatives to components that may not be in stock or
discontinued in the future. Users with access to editable CAD
files can tweak the part designs on their end to accommodate
alternative components in such cases. This allows the developer
to offload creating design variations to the users, who can
add compatibility for different COTS parts on their own, given
that the design files are understandable and the file repository
is easily navigable.
FABRICATION METHODS SELECTION
One of the main hurdles that open source hardware has to overcome
is that it requires fabrication of physical components to
reproduce the hardware. So, the fabrication methods that developers
choose for their components can have a significant impact
on the accessibility of their ORH project. Developers should
identify the requisite processes in parallel with the design of the
components and modify any parts that might be hard to manufacture
or source. Many ORH projects heavily rely on additive
manufacturing approaches such as fused deposition modeling
and more modern techniques like stereolithography, polyjet, and
digital light processing. These are quite approachable today,
even with desktop 3D printers, and are generally appropriate for
small parts that may have complex geometries. The range of
materials that can be 3D printed has grown considerably in
recent years [130], but most of these printed polymers are not
suitable for very high-force or long-duration applications. Metal
3D printing is starting to become more accessible, particularly
through on-demand prototyping services, but it is still limited in
terms of material choices and is sometimes cost prohibitive
[131]. There are also several other fabrication techniques that
use 3D printing as one of its steps. Printed parts can be used as
molds for different resins and are commonly used to make components
for soft-elastic actuators [132]. Other techniques have
taken the idea of molding with 3D-printed parts even further,
such as shape deposition manufacturing [133], which alternates
deposition and removal of material to create embedded structures
with different materials. HDM [22] builds on this concept
by including both permanent and sacrificial parts in the mold
and has been used in making the robot's grippers and hands of
the OpenHand [23] and OpenBionics [29] projects.
For larger parts, developers can consider subtractive methods
such as laser cutting, laser etching, CNC routing, and
waterjet cutting. These are often very fast, inexpensive, and
can work with a variety of materials. Although they are limited
to extruded planar geometries, many projects have found ways
to fabricate components using these methods. Piccolo [119]
and an early HapKit version [121] use mostly laser-cut acrylic
or fiberboard, the legs and frame of the quadruped Stanford
Doggo [75] are made with waterjet aluminum sheets, and the
Pupper [76]. The skeleton of the OPSORO robot [56] is also
mostly composed of laser-cut foam pieces that can be snapped
together. These pieces are also designed with assembly errorproofing
in mind, for example, with connectors of different
widths that fit together only in a specific way. Another lasercutting-based
technique stacks multiple layers of cut 2D parts
to create thick structures, as seen in WoodenHaptics [122]. The
OpenRoACH robot [83] is made from etched wooden sheets
with flexure joints that serve as creasing patterns so that the
2D sheets can be folded to create 3D robot structures. Lasercut
parts have also been used as stencils for painting different
dot patterns onto latex sheets in the Punyo tactile gripper [113].
Even without laser cutters or waterjet cutters, rigid sheets of
foam, plastic, cardboard, fiberboard, or other material can be
scored and cut with common tools to easily construct structures
for robots. Fabrication of the BigANT robot [84] presents
a versatile technique (called a PARF fabrication) of building
robots with rigid plate materials connected by fiber-reinforced
tape joints. These types of versatile fabrication methods that
rely on minimal tooling and less expensive materials can help
ORH projects reach a much wider audience of users.
Depending on the functional requirements of the components
in a robot, certain material properties may be desired,
which then inform the choice of fabrication method. Soft
robots use a number of different approaches to manufacture
elements of their robotic systems [134]. Many social robots are
also designed with soft, deformable materials because they
need to be compliant to external contacts, or blend with household
objects. The Blossom [53] robot's exterior is made from
knitted fabrics that are used to convey warmth as the robot
is intended to be stationed in people's homes. The OPSORO
robot [56] also uses flexible textiles as well as foam patterns
that are stitched together to make the 3D shell. Just as some
robots need highly deformable components, others may need
high-strength components to sustain large loads. The OSL [72]
is one such project that requires several machined metal parts.
But machine tools like CNC mills and lathes might be out of
reach for many users who may not be trained to use them. To
mitigate that, the OSL project offers the option of purchasing
a prebuilt leg or suggests that users outsource the fabrication
to machine shops. When possible, developers should still
try to recommend easy-to-use fabrication methods for their
components. And if no existing method is suitable, they could
rely on COTS components such as extruded frames, tubes, or
patterned plates, which can be readily adapted into large-size
robot builds. For example, FarmBot [120] and HOPPY [71]
both use metal extrusions; linkages of the PARA robot arm
[50] are made from acrylic tubes, ROBEL [31] is mostly composed
of off-the-shelf brackets, and Zumy [91] and HOPPY
are mostly made from COTS parts.
Even for users with no access to or expertise with manufacturing
equipment, obtaining custom-fabricated parts is
becoming easier with the rise of on-demand prototyping and
manufacturing services [135]. Developers could utilize the
wide range of manufacturing options available with popular
SEPTEMBER 2023 IEEE ROBOTICS & AUTOMATION MAGAZINE
141
IEEE Robotics & Automation Magazine - September 2023

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - September 2023
