Before, the robot had a couple of basic algorithms to identify objects via the cameras in its hands. As Eckert told us:
As you showed the robot an object, it took pictures at different angles and after users chose the best view, that's what it looked for in performing a task. Now, we've given it a number of different recognition algorithms to use, such as blob versus background, certain textures, certain shapes, hard edges, and irregular-shaped objects. You're training it to use one algorithm and one image per task, instead of building up a library of images. To show Baxter an object and pick one takes only 30 seconds.
All of this is in keeping with the main purpose for creating Baxter in the first place: an industrial robot for manufacturing processes that change a lot. A key attribute is to be able to quickly retrain the robot from one task or object to another, for the flexible and short-run manufacturing, or job shop manufacturing, done by small and medium-sized business that haven't had robots before.
In February, when I got to see the brand new Baxter at the MD&M West show in Anaheim, Rethink's intent was to offer an SDK this year to help open up the robot as a platform for development. To some extent, that occurred with the Baxter Research Robot version we told you about. But that's a stripped down version software-wise. Rethink removed a lot of its proprietary software, although it continues to add more capabilities of Baxter into the SDK that comes with that research robot version, said Eckert. "We do still intend to offer users the ability to start programming on top of our software stack some time in the future."
William, thanks for the honest feedback. (I know I can always count on you for that!) I think it's a good point that big manufacturers have something to gain from a safe robot. It will be interesting to see if any decide to use Baxter once, and if, it passes their performance thresholds. I do know that even big manufacturers--not automotive and probably not aerospace--often have a few stations that don't require super-duper high-speeds.
Ann, OK, it got away from me that Baxter was not aimed for everybody. But those same features should also make Baxter more valuable for the small organizations as well. And it doesn't hurt ones career to get a large order from a big company. Many years ago my little brother sold computerized cash registers, and he wound up selling them to the "GAP" stores. He was very shocked when he discovered how many he had actually sold, since he had not been aware of how large the chain was at that time. On rare occasions a surprise can be really nice.
What others should be able to learn is how to have a robot be a safe working partner. That would be a very valuable attribute for any robot.
It is certainly correct that those particular new features will extend the usefulness of the robot, and thus the number of users. But for som of the larger companies it will need to be able to demnstarte a big advantage over the robots presently in use. So the marketing people will need to write very accurately and make it interesting at the same time. No matter how good it is, people need to know about it in order to want it. And it needs to have a total cost of ownership that is a bit less then the competition.
Liz, we've both written about this most interesting robot and I think it's definitely one worth following. Since you reported on the research robot, I wonder if you have any insight into how soon we can expect an SDK with Rethink's full software stack, instead of the stripped down version shipped with the research SDK? I think Rethink hasn't really decided yet, but if you have any insight we'd love to hear it.
Baxter is already much less expensive (by several times)than what's been available before in industrial robots. It essentially creates a new, lower-cost category that didn't exist previously, which was its inventors' intention. That said, I think far911 raises a good point: it will be interesting to see what competition arises and how long it will take to make less expensive robots. Presumably, that will take another combination of ingenuity and technology advances like the one that produced this shift.
I'd love to try one of these on our production floor, but I haven't seen any arm tooling that would work with our product, and the weight capacity might not be enough. But, definitely neat technology that I'd like to play with if I could convince somebody to buy one!
Thanks for keeping up with this story, Ann. These are great innovations, especially, as you note, the expanded vision and ability to pick up and hold objects. This really shows where the future of robotic automation is heading in manufacturing and what benefits it can have.
Focus on Fundamentals consists of 45-minute on-line classes that cover a host of technologies. You learn without leaving the comfort of your desk. All classes are taught by subject-matter experts and all are archived. So if you can't attend live, attend at your convenience.