IEEE Robotics & Automation Magazine - June 2013 - 10

COMPETITIONS

The DARPA Robotics Challenge
By Gill Pratt and Justin Manzo

ny news broadcast reveals that
Mother Nature and human
nature are not always on our
side. Shifting demographics
make society increasingly susceptible to
both natural and man-made disasters.
Densely populated coastal cities prone
to flooding and bad weather invite
heavy tolls from natural disasters. Manmade disasters come in many forms,
such as chemical spills, weapons of
mass destruction, release of radiation,
and so on. Humanity must improve its
defenses against such threats, i.e., we
must prepare for the unexpected.
During the Fukushima disaster,
many Japanese citizens, aware of the
development of humanoid robotic platforms, asked, "Can't robots help us?" It
turns out that they could not, at least
not then. The Defense Advanced
Research Projects Agency (DARPA)
Robotics Challenge (DRC) aims to
work up to robots that can help in the
near future. The DRC is a program in
human-scaled robotics for disaster
response. DARPA's goal is to create
robots that could interface with human
environments, use human tools, and be
commanded by humans without specialized training. We want robots that
can adapt to assist in any disaster.
Why a Challenge?
DARPA issued a challenge to seek solutions to a series of problems by posing
them to the community. How can we
Digital Object Identifier 10.1109/MRA.2013.2255424

Date of publication: 6 June 2013

IEEE ROBOTICS & AUTOMATION MAGAZINE

june 2013

design a robot to respond to an unstructured set of tasks? How can we make it
work in partnership with a human
leader despite limited communications?
How do we benchmark performance?
What tools do we use? We don't know
the answers, but we do know that no
corporation, school, or individual is
going to solve these problems alone.
We also chose the challenge model
as it offers the best way of engaging
the global community. If the prize
structure is properly tuned and the
solution is valuable enough, we'll
reach not only major robotics contractors but also individual thinkers, hobbyists, and academics.
We felt a more diverse funding
model was necessary compared with
the previous model of the DARPA
Grand and Urban Challenges as we did
not have an existing technology base
(i.e., automobiles) from which to start.
We are providing development funds
to some teams, chosen through proposal evaluation, with a number of
down-select points based on performance. We are allowing other teams
with independent funding to compete
without submitting proposals. We will
remove a barrier to entry for software
developers by providing a robot and
technical support to teams that perform best in a preliminary, simulationbased competition; simulations will
run on an open-source tool that we are
developing in parallel and hope will
become a community-standard
resource for producing and sharing
new robots and testing environments.

In total, more than 100 teams from
around the world are registered: large
contractors, small businesses, universities, government agencies, and groups
of individuals.
Collaboration Brings Solutions
We believe that we can achieve unique
knowledge transfer through a community-wide challenge. For example,
as we develop the government-funded
platform and simulation environment, we have a hotline to experts
with a vested interest in the program's
success. If we are having difficulties in
developing a real-time simulation
environment or scoping requirements
for a walking controller, we can reach
out to hundreds of participants. What
elements are critical in rolling out
these tools? The decision should rest
with the engineers and scientists who
will use them and who collectively
possess the theory and applied experience to prescribe solutions.
By proposing a challenge that
addresses manipulation, locomotion,
perception, and communications, we
force experts in each domain to pool
their intellectual horsepower. We are seeing super-groups formed from institutions that might never have collaborated
otherwise. Between teams, we see an
increase in cross-talk using challengeendorsed forums; such cooperation
might not occur in a conventional, performer-based program. Nobody wants to
give away the secret sauce, but in pointing each other to key resources, teams
significantly raise the knowledge floor.

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - June 2013