911-strike.com     A wakeup call for non-violent political action

New Blog!  www.crookedshepherds.wordpress.com 

New pentagon overflight evidence: www.thepentacon.com 

The Five Sided Fantasy Island -- an analysis of the Pentagon explosion

Rebutting "Pentagon 9/11 Getting the Facts Straight"

Eyewitnesses and the Plane-Bomb Theory

Home
Up

Solar power
Peak Oil
Robotics
Space

 
 

Machine Vision and Robotics

Neural network techniques are widely used in machine vision for applications in manufacturing process automation. Existing techniques are excellent for checking surface quality and texture and for checking correct assembly. In the electronics industry, both optical and x-ray vision systems are used to check lead bonding and solder quality. Robotic systems are widely used in industry as well; for example, most cars today are painted by robots, which have excellent repeatability and can tolerate uncomfortable and toxic working environments.

There several of the commercially available machine-vision software packages, including Cognex and Amerinex. Open-source machine vision software is also available; see CVIPtools and and Vista.

Unfortunately, today's robotic systems are relatively limited in terms of flexibility and autonomy. Most practical factory systems are programmed for specific purposes. Future research and the application of advanced neural network design concepts should provide for increased capabilities in the future.

Robotics links

Here are some links to labs that are applying the new ideas in neural networks to problems in robotics. The Koch lab at Caltech has a vision system with focal attention:

http://www.klab.caltech.edu/

Another project is by Rajesh Rao at Salk Institute:

http://www.cnl.salk.edu/~rao/

Here are some groups doing very advanced work in robotics, although their algorithms may not involve direct implementations of generative neural nets:

Brooks lab at MIT
hand-arm manipulation at MIT

CMU robotics page
UCSD robotics page
Marthabot
 

An exciting interview with MIT professor Hans Moravec on the future of robotics can be found at:

http://www.pbs.org/wgbh/nova/robots/moravec.html

Moravec predicts that within 50 years, robotic systems will achieve human-level capabilities.

Most major universities and many companies are doing work in robotics. An excellent directory site is maintained by the University of Massachusetts:

http://www-robotics.cs.umass.edu/robotics.html