new boston-robotics dog / mule robot video
new boston-robotics dog / mule robot video
by limor
Tue Mar 18, 2008 5:10 pm
Published on Gizmodo , the part of the video that it slips over an oil spill has not been seen on any of the Boston Dynamics videos before. I have yet to see anything like this robot development in terms of looking so much like a living creature.
This is really the direction i would like to take the hobby humanoids - 100hz feedback loop, no pre-recorded gaits, continuous refining of the neural-net/ Bayes decision tree..
This is really the direction i would like to take the hobby humanoids - 100hz feedback loop, no pre-recorded gaits, continuous refining of the neural-net/ Bayes decision tree..
phpBB [media]
by NovaOne
Tue Mar 18, 2008 11:10 pm
This is really the direction i would like to take the hobby humanoids - 100hz feedback loop, no pre-recorded gaits, continuous refining of the neural-net/ Bayes decision tree..
Wow, definitely,
but where can I sign up, and what skills do we need to contribute? This is what I get when I Google this sort of thing....is it along the lines you are thinking of?
http://www.robocup.de/AT-Humboldt/publications/pdf/rc07-simloid-evo.pdf
Its a major paradigm shift, doesn't it deserve a new forum section eg "Methods of biped control without pre-recorded gaits" or "Neural algorithms for biped locomotion"?
by limor
Thu Mar 20, 2008 12:49 am
basically the idea is to feed to a powerful processor all the servo and sensor readings at 100 times/sec and let him update the servos at 100 times/sec.
what he does and how he records, learns and controls the servos is up to anyone's programmatic imagination. ( here's a simple example: http://cognitoresearch.com/Balancing_Robot.html )
this feedback control loop is the baseline for developing robot control algorithm. it is how control algorithms are developed for anything from industrial robots to autopilots.
unfortunately this control design/development is not handed over on a silver platter by the RC servo world where these humanoids inherited their DNA.
On the Bioloid forum there are some people who have attempted in creating closed loop control using Matlab/Simulink friendly development environment.
what he does and how he records, learns and controls the servos is up to anyone's programmatic imagination. ( here's a simple example: http://cognitoresearch.com/Balancing_Robot.html )
this feedback control loop is the baseline for developing robot control algorithm. it is how control algorithms are developed for anything from industrial robots to autopilots.
unfortunately this control design/development is not handed over on a silver platter by the RC servo world where these humanoids inherited their DNA.
On the Bioloid forum there are some people who have attempted in creating closed loop control using Matlab/Simulink friendly development environment.
