AAAI-11 Small-Scale Manipulation Challenge: Robot Chess

Fourteen years ago Deep Blue defeated the world chess champion. Yet, it was humans moving the pieces! Fine, dexterous manipulation in uncertain environments remains an open challenge for robotics. Motivated by this challenge we invite roboticists to face off in the 2nd Robot Chess Competition. Winning teams will be the ones that show the greatest skill at perceiving the world state, successfully manipulating chess pieces and recovering from real-world interference and uncertainty.

The AAAI-2011 Small-Scale Manipulation Challenge focuses on perception and dextrous manipulation of small, light-weight objects, so that robots of any size can compete. This year the domain will again be chess. There will be two divisions: a mobile manipulation division for robots that can navigate on a tabletop, and an unlimited division for fixed-base arms and human-scale mobile platforms.

For basic info, read the AAAI 2011 Robotics Challenge Events announcement.

Rules will be similar to last year (see below). The precise rules will be negotiated among the registered participants.

Letters of intent are now due by April 15, 2011 (the deadline has been extended). Qualification materials are due by April 26.

To submit a letter of intent visit the AAAI 2011 Robotics Program EasyChair page.

The Competitors

Carnegie Mellon University Tekkotsu Lab: Chiara robot
Georgia Tech "Golem Chesster": Schunk arm
Intel/University of Washington Custom built "Gambit" arm
University of Alabama iRobot Create, Lynxmotion AL5D arm, Zii EGG computer

Videos

More Videos:

• CMU: Chiara vs. Chiara; more videos at the Chiara chess page

• Intel/University of Washington "Gambit" arm: part 1, part 2

• CMU vs. Georgia Tech: part 1, part 2, part 3, part 4

2010 Rules:

• Chess board and chess piece colors may be determined by each competitor
• Play with non-standard board (standard 2.25in or slightly larger 2.75in squares)
• Board will have a contrasting color border around it that separates the board edge from the table
• Standard *chess* pieces (recommend Staunton standard tournament pieces No 5 - available at Amazon)
• Board with 5in squares with *chess* pieces in starting positions
• 5 inch neutral play area around board of infinite height (no positioning above board when not in play)
• Robot can be placed anywhere outside of play area at beginning of move
• Well-lit but arbitrary lighting
• Colors and first move chosen by coin toss

• No size limitation
• Can be controlled off-board (secondary computer)
• May utilize a persistent power source (but cannot disturb pieces on board)
• Must accept game play state via a network signal (to be defined)
• States are Ready - Beginning Sync state that allows each robot to survey the board and reset data structures , Go-period during which board move is made , Opponent-opponenet is making its move, Done-the game is complete
• No other user input may be incorporated

• 2 minutes to make move and clear play area without penalty; five minutes results in forfeit of move and manual removal of robot
• A game is 10 legal moves on each side (in the interest of time)
• 3 forfeited moves forfeits the game
• Pieces must be placed completely within a legal square, and no other piece may be left in an illegal position. Violations may be corrected by a human referee with a point penalty.

Winner of this level has the highest score. Scoring criteria include:

• 150 points for each successful move within time allotment
• 10 deduction for each piece that must be repositioned by referee
• 25 points for each piece captured

• Verbally announcing each move
• Verbally announcing the opponent's move
• Detecting mis-placed pieces

• Pieces start out in a random pile and the robot places them on the board
• Pieces are standing but otherwise randomly arranged, and the robot places them on the board
• Board starts out in an arbitrary mid-game configuration (so vision must be used to recognize piece shapes)

• Robot notices if a piece is knocked over, and picks it up
• Robot notices if a piece is knocked over, and asks for help

• Robot delivers running commentary on the game
• Robot only announces its moves, captures and win/lose