Presentation on theme: "Cognitive Computer Chess Comprehension or, How to Play Chess Against Something that’s Not Really There™"— Presentation transcript:
Cognitive Computer Chess Comprehension or, How to Play Chess Against Something that’s Not Really There™
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
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
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
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
Cognex MVS-8100 PCI Frame Grabber & software visioning system Captures 640x480 with external cameras Included Cognex Vision Library(CVL) for capture/manipulation of images
PatMax Most important piece of software included Recognizes visual patters to identify objects Can locate objects in a scene and give relative coordinates
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
PatMax Versatility, cont. Identifies objects despite changes in scale Measures objects with accuracy within 0.05% Successfully matches objects of random orientation
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!
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
Teaching Chess — Optics Everything can be boiled down to pattern recognition Recognition of each individual piece type/color Recognition of board states — what changed?
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)
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
Potential Pitfalls Optics accurate enough? Arm accurate enough? Optical system — will it work? Robot arm — will it be here?
Extra information Robot Arm system description and specification at Robot Arm system description and specification at ot-er-9.html ot-er-9.html Optical System specification at mvs8100.asp mvs8100.asp Ask Marc “EE” Marc