AAAI-98 Mobile Robot Competition

July 26-30, 1998, Madison, Wisconsin

Sponsored by the American Association for Artificial Intelligence (AAAI).

Co-chairs: Robin R. Murphy and Greg Dudek.

The mission of the 1998 competition is to serve AAAI, AI-robotics researchers, and the larger AI community by promoting innovative research through events which appeal to media and sponsors, while conducting these events in a format that facilitates comparison of approaches but at low risk to indivdual or institutional reputations.

The 1998 competition will support two events: one rewarding innovation in human-robot interaction, and the other challenging more traditional navigation, task planning, and mapping activities.

Differences from the 1997 competition

The 1998 competition is envisioned to have three major differences from previous years.
  1. In order to encourage bold technical innovation, the 1998 competition will make a series of Technical Innovation Awards , highlighting entries that have some noteworthy innovation regardless of how well the entry performed in the competition. This have been discussed in prior years, and were a well-received part of the RoboCup 97 competition.
  2. Following the 1997 competition organization, the human-robot interaction event will have an optional Technical Challenge Round , where the participants have the opportunity to demonstrate the ability of their robots to function in less constrained scenarios.
  3. Teams who receive travel money must attend and present at a one-day workshop following the conference, with the flavor of a AAAI symposium. All other participants will be strongly encouraged to attend and present. A research paper will be required one month in advance.
In order to be eligible for sponsored funding, competing teams may have to satisfy specific criteria related to their competitiveness and the feasibility of their approach. These may include: demonstrating a minimum level of functionality (as evidenced by a videotape of the robot doing basic tasks), intrinsic innovation, etc. We interpret this to mean that teams who start earlier and are transferring research results to their entry will get higher priority for funding.

Preliminary Description of Events

Event 1: Human-Robot Interaction Event (a continuation of the Hors D. event from AAAI 97)

Essentially this will be the same as last year, which was very well received by the non-robotics members of the AI community.

An important difference for 1998 is that each entrant is required to explicitly and unambiguously demonstrate interaction with spectators. This can take the form of speech synthesis, or other media. In general, the more natural and responsive to spectators, the better.

In keeping with IJCAI panel on The Next Big Thing, more natural modes of communication are necessary for society's acceptance of robots. Furthermore, this helps distinguish the AAAI competition from other competitions. Note that communication is not required for the Navigation event, but as in the past competitions, it will be rewarded.

Event 1: Technical Challenge Round

This event will require the robot(s) to be placed in three (or more) locations over a 2 day period in the conference hall and make announcements or pass out flyers to attendees. The robots will be expected to function autonomously for a minimum of 30 minutes (batteries should hold up that long!). There will be only 15 minutes for each team to configure the robot for new lighting and environmental conditions, to encourage more robust algorithms.

Event 2: Navigation (a continuation of the Find Life on Mars event)

The event will involve the scouting of an unknown environment containing obstacles. The event will involve (a) preliminary circumnavigation of the efficient exploration of the environment, (b) recognition of objects in specific locations, (c) returning a map showing the locations of each object (d) processing of a list of objects to obtain, (e) acquisition and return of selected objects. The objects will be indentifiable by color and/or shape and exhibit various degrees of discriminability. The simplest types of classification will be based on color while the most complex will be based on object shape: for example wrenches versus screwdrivers.

The exploration step (a) will not be scored explicitly, so long as it is accomplished within a predetermined time limit.

In order to inject a small element of reasoning, it may prove desirable to pose the specification of the set of objects to be obtained (d) so that it can be satisfied in various ways. This might allows systems to infer the best strategy for accomplishing the task. For example, if the object is to pick up 4 objects of the same color, different set of objects might be easier or harder to navigate to.

There will be only two categories , one for robots that emphasise manipulation and another for those that put more emphasis on "advanced" sensing. We do not expect to have a distinct category for multi-robot entries, however entrants will be explicitly rewarded for technical innovation.

Your Feedback

While the mission and the general framework have been set, everything else is open for discussion. We would like to have final scenarios in place by Jan. 1, and then finalize the categories for Innovation Awards and resolve any debates over scoring by the middle of May. To that end, we solict input as to the details of the rules, areas that should be considered for Technical Innovation Awards, and general suggestions to help us further the mission of the competition.

Send your comments to murphy@silverheels.Mines.EDU

last modified by Tucker Balch, 30 Mar 1998