The 3D printing of the parts is cool, but the overall implementation is not out-of-this-world awesome. Still, neat. Good job.
I recommend some more work on the training part. One method is to define the different positions you want the arm to be in, then have it interpolate the movements to go to the different position. Just from watching it move, it seems you're sending the servos each key-position, causing it to jump and not have smooth movements.
Really awesome. considering there is some company angling to make millions of dollars with this very notion (trainable robots), I think it's pretty impressive
As I understand it, ReThink Robotics' main selling point is to do with co-working (having a robot sensitive enough to work side-by-side with a human safely). The trainability is another point (but there are a lot more issues to deal with than poses : In particular, their robot must identify the position of the thing to be picked up by sight).
But the thing that is really disruptive (in the actual Christensen sense) is their price-point ($22k) - waaay below that of most industrial robots.
OTOH, while the Baxter is very cheap and flexible, at the moment it isn't very capable. But that will change, I'm sure. And that's the essence of a genuine disruptive technology.
Right, this is true. However, I think trainability is a more important point. As humans get phased out of manufacturing, co-working will be less and less important. However, simplifying the way that you train a robot -- letting someone who has the domain-knowledge move it's arms around and produce the correct functionality -- seems much more important, in the short and long term. I wonder if there are any other features of their robot that I am missing that could me be more important.
I agree, price is what's really amazing there -- and I'm assuming that this guy has created a prototype that can be scaled up, and I think it'll be even less than $22,000.
I recommend some more work on the training part. One method is to define the different positions you want the arm to be in, then have it interpolate the movements to go to the different position. Just from watching it move, it seems you're sending the servos each key-position, causing it to jump and not have smooth movements.