IEEE Robotics & Automation Magazine - March 2014 - 18

At the end of the summer of 2009, as
the summer interns left, we asked them
about their experience. While they enjoyed working at Willow Garage, they
all expressed frustration about how rapidly ROS code base was changing, and
how difficult it was to make progress
with the platform changing out from
under them. As a result, we embarked
on milestone 3-creating a release version of ROS with sufficient stability that
a community could depend on it for
their work (which would often extend
it). The entire software team focused on
hardening the code, producing automated tests, and, most importantly, creating tutorials that would enable people
outside of Willow Garage to use ROS.
Robotics Ph.D. students at Stanford
University collaborated on many of
those tutorials by giving valuable feedback. The success of this milestone was
a bit harder to see right away than the
first two milestones, but two indicators
were the rapid adoption of ROS outside
of Willow Garage, and the positive response of the 2010 interns to the same
survey question at the end of the following summer.
Milestone 4 was achieved in the
summer of 2010. The first batch of the
PR2 robots was completed, and the robots were shipped to eleven institutions
around the world. The institutions were
selected through an open call for proposals. We received 78 full proposals for
the PR2s, and attempted to choose the
ten that would best advance the state of
robotics by contributing back to ROS.
We failed to choose only ten, and in the
end, the PR2 beta program distributed
11 PR2s on two-year no-cost leases. (At
the end of the leases, all of the institutions found a way to keep their robots
permanently). In the fall of 2010, due to
significant demand from the academic
community, we began selling PR2 robots at a list price of US$400,000, but
with a 30% discount for labs that could
demonstrate significant contributions to
open source software.
In addition to the 11 robots that
were distributed in the PR2 beta program, Willow Garage kept a number of
robots for use in house. A series of
"hackathons" were held to celebrate the
18

*

IEEE ROBOTICS & AUTOMATION MAGAZINE

*

successful completion of milestone 4.
Each hackathon was one week long,
and intended to demonstrate that the
goal of creating a platform (PR2 +
ROS) that was powerful enough to
build significant robotics applications
quickly had been achieved. In one
hackathon, the team was challenged to
teach the PR2 to play pool. By the end
of a week, a team of seven engineers
had succeeded in getting a PR2 to make
some difficult shots, including angle
shots and some long shots. Another
hackathon was "Beer Me," which solved
the age-old problem of getting a cold
beverage without having to walk to the
refrigerator. Both of these hackathons
can be viewed on YouTube (search
"PR2 Poolshark" or "PR2 Beer Me").
This is not to say that all robotics
problems were immediately solved by
the PR2 and ROS. Milestone 2 demonstrated that a robot could open a door,
enter a room, and plug itself in, but the
doors could not have self-closing springs,
and the plugs had to be clear of any obstacles. More importantly, at a price point
of US$400,000, there were no customers
for pool playing or even beer fetching. As
a result, there was little interest in maintaining the demonstration code or extending it to address its shortcomings.
Nevertheless, the PR2/ROS platform was
successful for its intended audience and
goal. In research labs, graduate students
were now able to share code and demonstrations, and could do much more innovation than could have been done before
the platform existed.
Transition
In 2011, having achieved the initial goals
for personal robotics, we set out on a
new course, to get robots in use. We
wanted to understand if there was a
market for two-armed, mobile manipulators, and we wanted to take advantage
of some of the other opportunities that
had arisen during the PR2 program. We
began a program of systematically exploring market opportunities, educating
our team about entrepreneurship, and
ultimately spin-off companies. Internally, we began to focus on a much lower
cost platform that would harness the lessons learned with the PR2 and ROS to

MARCH 2014

serve the people in home or service environments, to be ready in 2014.
Scott Hassan saw huge market potential for remote presence technology.
It served a need people had-to be
somewhere else, avoiding travel or time
conflicts, and it was technically feasible
since it did not require autonomy. He
founded Suitable Technologies to commercialize the Texai prototypes we had
built internally, and spun off with about
seven engineers.
As we explored opportunities in the
home space, a particularly compelling
one arose to use the PR2 to help persons
with disabilities. Henry Evans, a mute
quadriplegic, saw a PR2 from the PR2
beta program on CNN, and contacted
Georgia Tech and Willow Garage to see
if he could use a PR2 as his external
body. Thus began the PR2 phase of the
Robots for Humanity project. Willow
Garage sponsored research in Charlie
Kemp's Healthcare Robotics Lab at Georgia Tech, and together with a team at
Willow Garage, researchers Kaijen Hsiao
and Matei Ciocarlie demonstrated that a
PR2 could be used in a home situation to
help Henry with some tasks. More importantly, the project gave Henry and
others in similar situations hope that in
the future robots really will make a difference in their lives. The Robots for Humanity project continues beyond Willow
Garage (see http://r4h.org).
Some of the technologies that were
developed during the course of building
applications with the PR2 and ROS became valuable even without a complete
ROS system running. One good example of this is MoveIt!, a package that
began as arm navigation (software to
move the arms through space), but encapsulated an entire library of planning
algorithms. Sachin Chitta identified the
need for this package and brought together a team of interns, researchers
(both at Willow Garage and at universities), and developers to make it a reality.
MoveIt! is part of ROS, but can also be
used independently of ROS, and is now
supported by SRI International. MoveIt!
is a good example of putting advanced
theory into practice, something we
strove for throughout the software systems we built.
http://www.r4h.org
Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - March 2014
