In fact it is a six degree-of-freedom industrial manipulator, and in the first step, I am going to use Programming by demonstration for deburring. Anyway, I am studying now about ML. Since I graduated in mechanical engineering, I have to first clarify the problem and define the area that I should work on.

I am working on a program that is a brain for a robot and learns all activities

http://www.intelligent-systems.com ar/intsyst/proposedBrain.htm

In the first step, I am going to Use PBD (Programming by demonstration) in order to learn an industrial robot for Machinery. Then I will go through deeper theoretical point in Application of ML in Robotics and its implementation.

Actually, Sarge, to use exact algorithm , it is needed a lots of programming for each task, so that is why it is proposed that we can program the robot for learning. Then for each different task, we just need to do some training which are not so hard and time consuming. In this case, also, we are going to integrate other previous work in Manipulator Control, such as force and position control, with learning part. So, some how, the learning algorithm will give us desired path with respect to some training set. Consider that the piece for machining will not have exactly the same geometry, I mean the production will have Tolerance in dimension and also some deviation, so the desired path will not be well define. In the first step, I am going to use GMM and GMR to define the appropriate Model of the Path.