Mission
The purpose of the Robot Challenge is to re-invigorate
the AAAI Robot Competition by ``raising the bar.''
We hope to present the community with a new challenge that
drives ongoing research and provides an effective public
venue for demonstrating significant new work.
Judges
We are fortunate to have a distinguished panel of judges this
year:
- Henry Hexmoor, UND
- Ian Horswill, Northwestern
- Leslie Kaelbling, MIT
- Sven Koenig, Georgia Tech
Task
In short, the task for the robots entering this challenge
is to:
Attend the National Conference on AI.
Yes, that's an ambitious
task, but that's why it's the Robot Challenge. We don't
expect anyone to completely accomplish the task in the next few years.
But we do
hope researchers will make significant strides in parts of the
task and share and build upon each other's contributions.
Registered teams
Registration
To indicate your intent to participate, please complete this
form. Alternatively, send your
questions and/or registration information by e-mail to
aaai-robots@gmu.edu.
Components of the task
Details of the task are given below, but you should not
consider this to be a hard-and-fast specification.
We would like to encourage broad participation, so if your
robot can do something interesting in the spirit of the
Challenge, let us know and we will consider including it.
Also realize that your robot may attempt just
a few of the sub-tasks, or perhaps a simplified version of the
whole task from end-to-end.
Safe, engaging (even entertaining) and useful
interaction with
humans is of growing importance in robotics research.
In that spirit
we envision a highly interactive event.
Here is a more detailed description of example components of the task
for a robotic attendee:
- Initialization at the front door of the conference center.
The robot will be pointed towards the entrance.
- Navigation to the the robot registration desk.
The robot will need to follow signs,
ask people, or devise some other means of
navigating to the registration desk. The intent is that
the robot should solve this step of the problem without
a map.
- Registration:
Upon reaching the registration desk, the robot may find
a line of people (and robots) waiting to register.
It should wait in line until its turn.
At this point, the robot will be provided a map of the conference
hall, a destination conference room and
a deadline by which to reach it.
- Navigate to the conference room.
The halls will be crowded with people (as at any conference)
and the robot will be faced with several sub-tasks along the
way:
- Take an elevator: the conference room may be located on
another floor, so competence
in elevator use (including button pushing), is a plus.
- Schmooze: some of the people crowding the halls are distinguished
researchers. Your robot should interact with them to help
advance
its developer's career. It can recognize the dignitaries by
face, (or in this first year by the colored badges they will be
wearing).
- Handle additional (volunteer) tasks on demand during the navigational
phase. Your robot may be asked to ``guard'' a room for a few
minutes or to deliver an object to another room.
The robot should only accept these additional tasks if
it can accomplish them and still reach the
conference room designated in step 1 on time.
It might also be necessary for the robot to prioritize
according to who asked it to complete the task as well.
- Reach the conference room on time.
- Finally the robot will give a two minute presentation
about itself and answer technical
questions (OK, so this is a long
range goal).
Details of the task will be tailored to highlight capabilities of
the robots entering, so input from potential participants is absolutely
encouraged. It is also perfectly acceptable for an entrant to target only
one of the multiple task phases listed above (e.g. schmoozing or navigation).
We realize most robots will need ``help'' in some
phase of the task, so ``short-cuts'' (e.g. mapping the conference hall
the night before) and limited modification of the environment will be
allowed -- but you must make sure the judges are aware of these
adjustments.
Driving technologies
Accomplishing this task will require the integration of a number
of developing and new technologies:
- Localization in a dynamic environment.
- Safe navigation in the presence of moving obstacles (people).
- Path planning (and replanning).
- Planning with time constraints and multiple changing goals.
- Visual tracking of people and face recognition.
- Voice and/or gesture recognition.
- Natural language understanding and generation.
- Effective and friendly tools for user-robot interaction.
Awards
Unlike past competitions, robots will not be ranked in this
challenge event. However, a number of awards for technical
innovation may be given for exceptional
demonstrations at the judges' discretion. Award categories
include, but are not limited to:
- Innovation in localization and navigation.
- Innovation in robot vision or sensor technology.
- Innovation in human-robot interaction.
- Innovation in real-time planning.
- Innovation in manipulation.
- Excellence in collaboration and integration.
If no robot demonstrates exceptional ability in
any of these areas, the judges may decide not to
bestow any awards. Conversely, if several robots
show excellence in one category, multiple awards may
be given in that category.
Written summaries
To ensure the technical advances
are clearly understood, each entrant should also
submit a four page paper summarizing their
approach. You will have an opportunity to
present your paper at a workshop towards the
end of the week. More information on format
will be provided later.