Interfacing RTB to real arm robot

[Peter Corke]

I've been pottering around with a hobby class robot (4 joints plus gripper servo), see attached picture, and have some MATLAB code to communicate with it and drive it.  The robot is reasonably inexpensive, under USD 400, which makes a robot lab at home feasible.

The attached document is quick brain dump of what I've done.  I'd be interested to correspond with anybody who's interested in following this line of work or extending it.  There are quite a few robots in this class with mega servos and typically less than 6 DOF, so hopefully a few of them could be supported with not too much work.

You can find the code in the svn repository

http://code.google.com/p/matlab-toolboxes-robotics-vision/source/browse/matlab/robot/trunk/RobotArm.m

http://code.google.com/p/matlab-toolboxes-robotics-vision/source/browse/matlab/robot/trunk/interfaces/Arbotix.m

peter

[Joern Malzahn]

Peter,

we are currently doing lab courses with an old 5 axis Katana robot. Since the number of students in our courses has been steadily increasing during the past years we are currently tossing around some ideas for replacing this lonely old dinosaur.

I wonder if a couple of Phantom X interfaced with Matlab and the Robotics Toolbox would do a good job for hands on teaching of programming and kinematics basics. Could you comment on that from your experience with the Phantom so far?

Thank's a lot.
Joern

[Peter Corke]

The Katana was a nice arm but pricey.  I haven't used the Phantom for teaching but I think it would do to show the principles and the price is good. The assembly is non-trivial, simple but time consuming.  In operation it's not as smooth as a "proper" robot, if you give it a big displacement it will attempt to go there in one time step, so there is perhaps need for some lower level trapezoidal motion control.  The most logical place to put this would be in the controller board, so far I just used the out of the box embedded software for that.  Maybe we could work on polishing up that code.

5 axes is a bit of a curse because general approaches to inverse kinematics won't work, but maybe in your class you teach this from first principles.  IK is the thing most people seem to struggle with…

Happy to chat more.

peter

[Joern Malzahn]

Peter,

Thank's for the infos. Sounds very promising but also a bit of work, too. I still quarrel with myself, if there will be a Phantom beneath my Christmas tree... I should hurry up, I know ;).

Yes, we teach IK from first principles. Same with joint control and trajectory generation. Currently we emulate a single robot joint using ECP 220 setups. But for hands on experiments in kinematics, it would be nice to have a bunch of robots.

Polishing the code to get RTB more in touch with real robots sounds exciting and a nice thing to do. I'll let you know when things are making progress over here.

Jörn