models. One approach to this problem
is to make 3D printing more portable. 35
Another challenge is rethinking CAD
in an object-centric fashion, meaning
that models would be designed with
respect to an existing object.
While attachment is a basic capability needed for many 3D printed objects
to be functional, it is only the first step.
For an end user, the design of the
function is at least as difficult as the design
of the attachment. Sample tools that
address this problem include Grafter, 35
RetroFab, 32 and Reprise9 (Figure 12).
Adaptation and reuse. Bridging the
gap between geometry and function
presents a substantial challenge, even
for experienced users. A powerful way
to bridge this gap permits the work
of experienced designers to be easily
adapted and reused by others. Many resources for 3D-printable designs have
been extensively studied; they show
that adaptation of existing designs is
often trivial but rarely improves on the
original designs. 29 However, CAD tools
and the models they produce, while
general and powerful, are not necessarily designed with reuse in mind.
Functional information implicit in
an object’s geometric form is never
expressed explicitly; hence, it is inaccessible to anyone who is not also sufficiently skilled to recognize the underlying mechanical rationale.
Modelers would benefit from the
equivalent of an end user programming tool and a set of abstractions for
encoding design information in an interactively accessible way. This is what
the Parameterized Abstractions of
Reusable Things (PARTs) framework
provides; it puts advanced methods
for capturing 3D modeling design
intent into the hands of non-expert
modelers. 14 Doing so supports reuse,
experimentation, and sharing.
speed up a print by reducing the
amount of printed material. 28 To im-
prove 3D printing interactivity, a way
to design for embedded electronics
is needed. 36 Jones et al. approached
this by combining sculpting and
modular interaction toolkits to proto-
type interactive sculptures. 21 Alterna-
tively, 3D printers could produce a
new facade supporting alternative in-
teractions for existing physical inter-
faces (for example, Ramakers et al. 32).
These examples do not specifically
address or provide control over how the
3D-printed object should be attached
to the real-world object it is modify-
ing. A set of attachment methods could
provide a basis for exploring and modi-
fying alternatives. Several challenges
arise when attaching objects:
Collision. If an object is on the print
bed (to be printed on or through), the
design of the attachment must ensure
no collisions occur between the print
head and object. A design tool can de-
tect and visualize potential collisions
to help the user determine a viable po-
sition for the attached component.
Insertion. Specifically when printing through an object, there must be
a viable insertion path for the object.
Such a path can be estimated using a
reverse gravity model (that is, if the
object can easily fall out when inverted, it can easily be placed when in
Figure 11. The Reprise workflow assumes the existence of a model of the object to be adapted.
Durability. Strength or durability of
the attachment can be influenced by
the size of the connection (a very small
footprint connecting two objects is less
secure), the object’s flatness, and the
direction and area of force applied to
the attached object.
Semantics. At a higher level, the intended use can influence the effectiveness of an attachment. For example,
balance, direction of hold (for a handle)
and cost might be concerns that influence an effective attachment technique.
Automated tools such as Autoconnect24 help address these challenges by
creating customized connectors,
which take into account the position and weight of the objects being
connected. Interactive tools such as
Encore system8 can support exploration of potential attachment techniques and visualize the effectiveness
of attachment over a possible set of
metrics (Figure 10). Further research
is needed to determine the best
metrics, and the best way to express
those metrics computationally.
An open area for future investigation is how to develop tools that function in real-world settings where an
object to be modified may not portable
or is too large to be brought into a scanner or 3D printer. This requires the
high-quality, low-cost capture of real-world object models and ideally the
ability to convert them to high-quality
Figure 10. Encore visualizes attachment goodness using a heat map. 8
Three different metrics are shown: (left) Viability for printing; (center) Likely durability based on curvature; (right) Estimated usability based on the assumption that balance will be better in areas near to the
center of mass (This assumes the forces applied have the same direction as the surface normals).
It starts with a specification of the type of action to be supported. Each action has a set of associated adaptations, from which the user picks. Reprise
generates an appropriate model adaptation. Parameters of the adaptation can then be adjusted. Finally, an attachment method is selected. 9