The team had just 10 months to design and construct the structure, which
included creating all of the code for the robot. Yet it still had to spend
time managing and aligning both public and private stakeholders, such
as the state’s forestry department and the companies manufacturing the
timber and robot.
“There were so many different parties involved, so it took a while for
everyone to get in sync,” Mr. Schwinn says. The team lost several weeks
trying to get its proposal approved. Schedule adjustments had to be made
to make up for that time.
“Because of the limited time frame, we moved the optimization stage to
the end of the project,” Mr. Schwinn says. “Now we’re evaluating how the
process unfolded and how the system behaves over time. Then we learn
from it and try improve the tools that we used.”
Training the Robots
The project team had to focus not
just on the design parameters of the
robot that would fabricate the intricate finger-joint pattern along the
perimeter of each plate. The team also
had to consider the design parameters of the project’s other digitally
controlled machines—such as the
Machine, which cut the
large plywood panels before
the robot created the joints.
The Hundegger also cut the
pavilion’s wood-fiber insulation and cladding layers.
“This ultimately meant
that we had to develop
similar programming techniques for those machines
more or less on the fly in order to be able to use
them with the same efficiency as we programmed
and used the robot,” Mr. Schwinn says.
Conventional modeling and control techniques
for digitally controlled machines used in the indus-
try usually involve a large
amount of manual model-
ing work and plausibility
checks, he says.
“Instead, we embedded all the modeling in
our rule-based algorithms,
which ensured that the
fabrication of all individual elements would adhere
to the same standards,” Mr. Schwinn says. “If we
hadn’t done that, this would have created a significant bottleneck that could have derailed the entire
“The result is
as beautiful as