1
Challenge Contact: Prof. Dr. Andreas Nüchter, +49-421-200-3181, a[email protected] Approach Results Future Work References Development of a Cheap Chess Development of a Cheap Chess Robot: Planning and Perception Robot: Planning and Perception Build a chess playing robot with as minimal constraints as possible on the game. Allow for a non-static camera setup i.e. camera movable around during the game. Use a standard chessboard set with no modifications. Use off-the-shelf camera, e.g. In our case, approx 50 Euro Logitec Webcam. Use a cheap robot arm, in our case a 4 DOF + 2- Plate Gripper arm with joints driven by AX-12 Dynamixel servos, all assembled from standard kit. Use a standard chess engine to generate the appropriate chess moves. Use image snapshots of the game scene at every time step and compare to determine moves made. Use image processing to find the 4 corners of the chessboard, perpective projection, line and Canny detection [2] etc. Pass moves to chess engine and interpret given moves into locations using chessboard size information. Use BiRRT [1] for motion planning and compare with just computing the end effector IK at each iteration. Add constraints into the planning like 'uprightness' of chess pieces. Implementation using OpenCV [5], OpenRAVE [4] and ROS (Robot Operarating System) [3] Experiments with perception showed reliable move detection with real-time perfomance comparable to previous modified chess cases. Motion planning produced better trajectories (smoothness) compared to manual IK solving techniques. Overall, performance in playing chess comparable to those of previous robots with modified chessboards/configurations. Adding more links to the robot arm to have 6 DOFs which are easier to plan with. Using a depth camera, e.g. Kinect for the move detection. [1] R. Diankov. Automated Construction of Robotic Manipulation Programs. PhD thesis, Robotics Institute, August 2010. [2] M. Sonka, V. Hlavac, and R. Boyle. Image Processing, Analysis, and Machine Vision. Thomson-Engineering, 2007. [3] http://www.ros.org/wiki [4] http://www.openrave.org [5] http://opencv.willowgarage.com/wiki/ Billy Okal and Oliver Dunkley Contact: Billy Okal, [email protected]

Development of a Cheap Chess Robot: Planning and Perceptionokal/docs/talks/gi_poster2012.pdf · Development of a Cheap Chess Robot: Planning and Perception • Build a chess playing

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Development of a Cheap Chess Robot: Planning and Perceptionokal/docs/talks/gi_poster2012.pdf · Development of a Cheap Chess Robot: Planning and Perception • Build a chess playing

Challenge

Contact: Prof. Dr. Andreas Nüchter, +49-421-200-3181, [email protected]

Approach

Results

Future Work References

Development of a Cheap Chess Development of a Cheap Chess Robot: Planning and PerceptionRobot: Planning and Perception

• Build a chess playing robot with as minimal constraints as possible on the game.

• Allow for a non-static camera setup i.e. camera movable around during the game.

• Use a standard chessboard set with no modifications.

• Use off-the-shelf camera, e.g. In our case, approx 50 Euro Logitec Webcam.

• Use a cheap robot arm, in our case a 4 DOF + 2-Plate Gripper arm with joints driven by AX-12 Dynamixel servos, all assembled from standard kit.

• Use a standard chess engine to generate the appropriate chess moves.

• Use image snapshots of the game scene at every time step and compare to determine moves made.

• Use image processing to find the 4 corners of the chessboard, perpective projection, line and Canny detection [2] etc.

• Pass moves to chess engine and interpret given moves into locations using chessboard size information.

• Use BiRRT [1] for motion planning and compare with just computing the end effector IK at each iteration.

• Add constraints into the planning like 'uprightness' of chess pieces.

• Implementation using OpenCV [5], OpenRAVE [4] and ROS (Robot Operarating System) [3]

• Experiments with perception showed reliable move detection with real-time perfomance comparable to previous modified chess cases.

• Motion planning produced better trajectories (smoothness) compared to manual IK solving techniques.

• Overall, performance in playing chess comparable to those of previous robots with modified chessboards/configurations.

• Adding more links to the robot arm to have 6 DOFs which are easier to plan with.

• Using a depth camera, e.g. Kinect for the move detection.

[1] R. Diankov. Automated Construction of Robotic Manipulation Programs. PhD thesis, Robotics Institute, August 2010.

[2] M. Sonka, V. Hlavac, and R. Boyle. Image Processing, Analysis, and Machine Vision. Thomson-Engineering, 2007.[3] http://www.ros.org/wiki [4] http://www.openrave.org [5] http://opencv.willowgarage.com/wiki/

Billy Okal and Oliver Dunkley

Contact: Billy Okal, [email protected]