Meet iRobot Founder Rodney Brooks's New Industrial Bot, Baxter

First time accepted submitter moon_unit2 writes "Technology Review has the scoop on a new industrial robot created by famed robotics researcher Rodney Brooks. The robot, Baxter, is completely safe, extremely adaptable, and ridiculously easy to program. By providing a way to automate simple manufacturing work, it could help make U.S. manufacturers compete with Chinese companies that rely on low-cost human labor. You can see the new robot in action in a related video of the robot in action and Brooks discussing its potential." $22 thousand and shipping next month, goes the story.

Comments

after watching the interview

[v1]

I wish they had showed some practical application. Moving an air hockey paddle two feet to the right isn't extremely practical. Show me it loading a dozen donuts into a donut box or something.

Sensors. Yes it has force sensors but anything else? He was having to carefully position the paddles for pickup. He talked about previous robots being "blind". But is this robot really not blind? Blind people have a sense of touch, why isn't this robot "blind"? Show me it can adapt a little using sight or ultrasound or something.

Slow. Wow. Ten seconds to move the paddle. Traditional industrial robots would do ten paddles in ten seconds. Sure they're not safe to be around running at those speeds, but this is completely at the other wrong end of the speed scale. Nobody's going to use a robot that moves like a retarded sloth. I do hope the speed can be cranked up?

I would like to have seen a very brief runthrough of the training process. Telling me ten times that it's "easy" without showing me it even once leaves me suspicious of your definition of "easy". (and of "simple")

Someone setting their hand under an object being set down really isn't a practical example of collision behavior on the manufacturing floor. Stick your head out in front of the arm's path and show me how it reacts. Does it knock you off your feet, or maybe shove you slowly to the side? Does it stop immediately and drop that fragile widget a foot down onto the bench? This demo wasn't nearly as informative as I was hoping it would be.

But I do like the "move the arms" training method. I'd put a little time into pondering how to train manufacturing robots in the past, and I was always wondering why they didn't use that approach, at least to rough out the behavior, and use an interface to tweak the positioning and timing etc. But afaik all the programming on other industrial robots to date has been purely through the console. Even if you don't eliminate the programmers or computer techs, at least being able to get a good floor worker to flesh out the robot's basic movements will save a lot of time. And if you involve them more, they can help in optimizing the behavior too I think.

Re:Slow Movement

[weiserfireman]

I work in a CNC Machine Shop. A robot like this would be great for unloading and reloading the Lathes for example.

I don't need it to be incredibly fast. It takes 2 or 3 minutes to run a part anyway. I just need it to be almost as fast as a person. If I can train it to pick up a blank, load it in the lathe, unload it when the cycle is complete, and stack the parts neatly in a tray, I free up a person to go complete a setup on another lathe, troubleshoot a process, complete an SPC chart, or go home and get a good nights rest while the robot runs parts for us.

It doesn't have to be fast, just fast enough.