1. Cognitive Computer Chess Comprehension or, How to Play Chess Against Something that’s Not Really There™

2. New Equipment Overview  SCORBOT-ER9  Industrial-Grade Robotic Arm  MVS-8000  Frame Grabber and Embedded Vision System  Shiny new toys; what they do, and what we’ll do with them

3. SCORBOT-ER 9  Robot Arm with 5 degrees of freedom  Can carry up to 2 kg (4.4 lbs.)  Position repeatability of 0.09mm (0.004”)  Speed of movement defined in mm/sec or 1% of max. range

4. Control Software  Controller Internal Language, ACL (Advanced Control Language)  Interface to ACL controller provided via ATS interface software  SCORBASEpro — Windows-based software for control  Hand-held “teach pendant” allows direct manipulation of arm

5. Better Control Software  RoboCell — full compatability with SCORBASE programs  Simulation software for arm and objects  Direct interaction for position teaching  3D graphical display for simulation  Allows for offline manipulation and testing

6. Cognex MVS-8100  PCI Frame Grabber & software visioning system  Captures 640x480 with external cameras  Included Cognex Vision Library(CVL) for capture/manipulation of images

7. PatMax  Most important piece of software included  Recognizes visual patters to identify objects  Can locate objects in a scene and give relative coordinates

8. PatMax Versatility  Fault tolerance in PatMax allows it amazing versatility in locating and recognizing objects  Despite surface irregularities and reflectivity, recognizes semiconductor wafer in varying conditions

9. PatMax Versatility, cont.  Identifies objects despite changes in scale  Measures objects with accuracy within 0.05%  Successfully matches objects of random orientation

10. What to do with the system?  First, work on basic motions/pattern matching  Then, start putting things together to make a more complex system  Start with Chess!

11. Chess and the Robotic Arm  Goal — to be able to play chess against the arm  This will allow us to grasp the abilities of the system while leaning how to use it  Also, it’ll be really cool to be able to play chess against a robotic arm

12. Teaching Chess — Optics  Everything can be boiled down to pattern recognition  Recognition of each individual piece type/color  Recognition of board states — what changed?

13. Teaching Chess — Mechanical  Each time the arm has a move, it needs to accurately move a specific piece  Identify each square on the board by relative coordinate system  Need to measure distance to desired coordinates  Use well-known geometry to calculate how far to move in what direction  Capturing opponent’s pieces  Hit the clock (known location)

14. Teaching Chess — Fitness Algorithms  Many different kinds of fitness algorithms are around  Much discussion in Russell/Norvig  Marc and I consider the choice and construction of a fitness algorithm to be trivial  Left as an exercise to the audience members

15. Potential Pitfalls  Optics accurate enough?  Arm accurate enough?  Optical system — will it work?  Robot arm — will it be here?

16. Extra information  Robot Arm system description and specification at  Robot Arm system description and specification at http://www.intelitek.com/products/robotics/systems/scorb ot-er-9.html http://www.intelitek.com/products/robotics/systems/scorb ot-er-9.html   Optical System specification at http://www.cognex.com/marketing/products/prod_8000_ mvs8100.asp http://www.cognex.com/marketing/products/prod_8000_ mvs8100.asp   Ask Marc “EE” Marc