Even though I knew what to expect, seeing Baxter in action here at Automate 2013 was still amazing. Baxter is Rethink Robotics’s inexpensive robot designed for continual operations in material handling applications. Physically, he looks like something out of the Jetsons: a thin center frame with two joined arms that end in small claws, with a small, iPad-like screen hovering where you would expect the head to be. And, sure enough, Baxter’s screen displays a cartoonish face, focusing on the operation at hand.
But Baxter represents a type of technological leap that once seemed impossible.
Baxter requires no integration and no programming. In fact, Baxter doesn’t even need to work with a specialist. Because of “his” advanced software, Baxter is able to learn kinetically through a simple series of movements and an advanced graphic interface and vision system.
In the demonstration, a volunteer from the crowd — sadly not me — was chosen to “teach” Baxter’s right arm, while the Rethink engineer “taught” the left arm.
Several small ceramic ramekins were placed in front of Baxter. The engineer guided the left arm toward one, lowered his claw into it, and pressed a button. “Baxter will tell me when he’s confirmed that he’s learned what to do,” he said. Sure enough, the little screen atop his shoulders that displayed a face nodded.
Then, the engineer physically moved Baxter’s arm up, over and down, placing the ramekin further down the table. Another button press, another nod, and Baxter knew what to do. He then moved his arm back to the starting spot and grabbed the next ramekin in the pile, moved it up and over and placed it down on top of the first.
The audience volunteer followed along, and soon the arm he was responsible for had learned how to move ramekins across the table, as well.
The engineer next demonstrated safety features, putting himself in Baxter’s way to show that Baxter could sense other objects (or humans) and cease movement to prevent injury. Baxter is also “self-aware,” but not in a scary SkyNet-from-Terminator way: he is also incapable of ramming one arm into the other and damaging himself.
Additionally, when Baxter ran out of new ramekins, he moved his claw around where he expected to find one and then made a “face” — one of his eyebrows rose while the other curled down in concern. Baxter will stop working if he senses an operational problem, too.
The entire demonstration lasted a few minutes, and this involved answering questions from the crowd and explaining the whole process. It’s surprising to think how quickly Baxter can learn new tricks, even complicated material handling and supply chain operations.
This type of technology is gradually improving in robots, and is available in top-shelf robotics systems as well. But this type of technology in an affordable robot broadens the availability of robotics to small- and medium-sized companies, which have previously been unable to afford such technology. When anyone on the line can “teach” a robot how to commit itself, the entire staff is suddenly an expert on process. And considering that Baxter can learn materials handling and supply chain operations, how far are we from robots that can learn more complex operations?