1. Final Presentation Advancing the soccer robot Ben Jenkins Supervisors: Gordon Lowe, Dr. Haim Hiok Lim Mentor: Charles Greif

2. Overview Background Objective This years work Further development Additional information

3. What is the soccer robot Autonomous vehicle Plays soccer Development at Monash Uni started in 1999 My research topic for Hons RoboCup hosts soccer competitions

4. RoboCup Worldwide research and education initiative Foster AI and intelligent robotics research Provides a standard problem where a wide range of technologies can be used. Project-oriented education.

5. Building blocks Sensors Field Positioning System Motion Artificial Intelligence

6. Objective To be able to make the robot “see” the ball and be able to move towards it.

7. This years work Revamped design computers. Built cable for motion system. Built frame for camera, motherboard and kicker. Developed preprocessing filtering for vision system. Built application for determining the real distance between robot and ball from an image. Developed code to communicate with the motion control card and servo boards. Convert the motion code from windows to QNX.

8. This years work (cont) Integrate new motion with previous modules Test and debug

9. Motion Control System Built control cable from computer to the servo control boards. Develop software to communicate with MCC and SB. (Made wheels turn). Further development to make the robot move. Develop routines to actuate the kicker. View cableWheels spinningRobot movingNext SlideKicker

10. MCS (cont) Implemented the quintic polynomial control system. Allows displacement, velocity and acceleration to be defined at the start and end points of the robot trajectory. Software developed, but not fully tested. Simulations appear to gives correct results. MATLAB graphsNext Slide

11. Control Cable Return Computer Motion Control Card PCI bus Motor 1Motor 2Motor 3 Servo 1Servo 2Servo 3

12. Wheel spinning Return

13. Robot moving Return

14. Pneumatic kicker Return

15. MATLAB simulations Return

16. Vision system Optimised previous code. (40% of the size). 2 Processes Read from grabber. Write to BMP file. Send message. Read BMP from file. Process the image. Delete file. Existing system Single process Stores image in memory (passes memory pointer). New system

17. Message passing Totally rewrote the message passing class. Process was writing globally hoping the intended receiver was listening. Now use QNX message passing functionality. Process ID’s are stored and messages are addressed.

18. Distance mapping Image needed to be processed to determine direction and distance of ball. With the assistance of Stewart Hore, we rebuilt an application which would read in images and provide a series of co-efficients to hard code into the source. Filtering was required. 3x3 was good. Could afford 5x5. Filter imagesNext Slide

19. Filtered images Original3x3 mean filter5x5 mean filter Return

20. Further development Integrate motion and vision system. Full control over robots movements Make robot kick ball Make robot work as team member Enhance vision. – Use better camera. Digital image. Enhance motion – 400mm/sec max speed – 2500mm/sec to be competitive

21. Additional information bwjen1@csse.monash.edu.au www.csse.monash.edu.au/~bwjen1 RoboCup website – http://www.robocup.org/02.html Gordon Lowe, Bldg 75, Rm 1101, Clayton.

22. Any questions ?