Figure 9. Some examples of objects designed for measurement uncertainty.
Figure 7. Some examples of inaccurate measurement practices. 23
(a) The tick mark on the paper is not aligned with the end of the phone for measuring phone length.
(b) A ruler is not an ideal way to measure angles accurately. (c) The width of the light bulb’s base
is difficult to estimate, and the task is ill defined (that is, should the outside or the inside of the threads
(a) This tripod’s angle can be adjusted. (b) Part of this handle can be replaced to resize it. (c) This cup
holder has buffers in it. For most of these objects, flexibility in the face of error also affords new flexibility in use (for example, the cup holder can fit many cups). 23
(a) (b) (c)
code information specific to reuse to
be easily modified for new contexts. We
discuss each challenge.
Measurement. When a model must
conform to a specific real-world goal
after it is 3D printed, it is important
that the goal be precisely specified.
However, measurement errors pose a
significant yet often overlooked challenge for end users, as determined by
a systematic study of the sources and
types of such errors. 23
Kim et al. found that user error
(such as misaligning instruments and
misreading units), measurement instrument precision, and even task
definition made measurement error
common. 23 Figure 7 depicts some examples of faulty measurements from
the study. Compounding these errors
is the fact that 3D printing itself is not
perfectly precise. For example, some
materials shrink slightly as they cool.
Thus, measurements are at best approximations that contain some degree
of uncertainty. A model robust to this
uncertainty will be less likely to fail.
Measurement error can be addressed at the prototyping stage using
mixed design approaches that incorporate simple materials such as foam
or Lego Bricks.® 28 This same approach
has been successful at facilitating experimentation and iteration in AT design. 15 An example is our iterative design of the cello bow using Lego Bricks
to estimate length (Figure 8).
Design strategies can accommodate measurement error, as illustrated
in Figure 9. For example, by inserting
a flexible buffer around an uncertain
real-world object, small differences
would no longer require reprinting. A
related (and synergistic) strategy could
support the replacement of small areas
of a 3D-printed object likely to have errors. This would reduce cost and waste
because that region could be reprinted
and then connected with a snap joint,
adhesive, or other method. Innovation
is needed to further expand this set
of methods and develop robust automatic tools for applying them in a wide
range of contexts.
Attachment. For functional ob-
jects to be useful, they must typically
interact in some way with real-world
objects (people or items to be ex-
tended, manipulated, or repaired).
Interaction, in turn, typically requires
attachment, the temporary or perma-
nent connection of two or more ob-
jects. Thus, the problem of attaching
3D-printed object to a real world one
must be addressed.
Attachment has been explored ex-
tensively outside the domain of 3D
printing. Material properties, strength,
usability, and aesthetics must all be
considered when attaching objects.
The issue is sufficiently complex to
support websites such as ThisToThat
(Glue Advice),a which help answer
questions about how to connect two
objects with glue.
In the domain of 3D printing, incorporating existing objects is also important. Incorporating Lego Bricks can
Figure 8. Prototyping the length of a cello bow holder (inset shows final result). This length
was challenging to determine due to the lack of a physical object to measure and physiological
subtleties in finding the right length for the dynamic activity of cello playing. 15