About two months ago, a new employee arrived on the production line at Vanguard Plastics in Southington, Connecticut, a town that was once a hub of U.S. manufacturing but saw many of its factories disappear in the 1960s. The small manufacturer’s new worker, Baxter, is six feet tall, 300 pounds, and a robot. For a hulking machine, Baxter is remarkably expressive. A pair of eyes on the screen that serves as a face stare down as the robot picks up plastic components, look concerned when it makes a mistake, and direct its glance at its next task when one is finished. It’s cute. But the real point of these expressions is that they let workers nearby know instantly if Baxter is performing appropriately, and they provide clues to what it is about to do next. Even more amazing, when Baxter is done with one task, a fellow worker can simply show the robot how to start another. “Almost anyone, literally, can in very short order be shown how to program it,” says Chris Budnick, president of Vanguard Plastics. “It’s a matter of a couple of minutes.”
Baxter is the first of a new generation of smarter, more adaptive industrial robots. Conventional industrial robots are expensive to program, incapable of handling even small deviations in their environment, and so dangerous that they have to be physically separated from human workers by cages. So even as robotics have become commonplace in the automotive and pharmaceutical industries, they remain impractical in many other types of manufacturing. Baxter, however, can be programmed more easily than a Tivo and can deftly respond to a toppled-over part or shifted table. And it is so safe that Baxter’s developer, Rethink Robotics, which loaned Baxter to Vanguard Plastics, believes it can work seamlessly alongside its human coworkers.
Baxter’s talents could, for the first time, bring the benefits of robotics and automation to areas of work where it never made sense before. This might mean lost jobs for an already struggling low-skill workforce. But it could also help the United States compete in the global manufacturing market against countries that offer low-wage labor. The amount of manufacturing in the United States has declined dramatically over the last few decades, in part because workers in other parts of the world will perform simple and menial jobs for far less. But robots like Baxter could wipe out that advantage by taking over such tasks, making it once again possible for many industries to competitively manufacture their products in the United States and other developed countries.
Perhaps it’s unsurprising that such an innovation should come from probably the biggest name in the world of robots, Rodney Brooks. During the 1980s and 1990s, while at MIT, Brooks published work that helped shape the twin fields of robotics and artificial intelligence. Then, in 1990, together with two of his students from MIT, he cofounded iRobot, a company that introduced robots to new fields of work. It now makes a wide range of domestic and military robots, including the Roomba, a Frisbee-size household vacuum cleaner that has sold more than eight million worldwide.
The idea for Baxter came about, Brooks says, while iRobot was searching for a suitable manufacturing partner for its one of its products. He remembers marveling at the fact that so much electronics manufacturing was still done by hand and how much of this kind of manufacturing had moved to low-wage economies in Asia. “I thought, ‘Are we going to be doing this in 500 years—still chasing cheap labor? There’s got to be a different way,’” he says.
Brooks, who founded Rethink in 2008 after leaving both iRobot and MIT, realized that robots could take over many jobs if they could be made safe, adaptable, and ridiculously easy to program. “Everyone was thinking about how to do it with the current industrial robots,” he says. “I said, ‘Let’s make a different sort of robot for industry.’”
The result is something very different indeed. Much like a human worker, Baxter can be taught in minutes how to recognize a new object or perform a new task. To teach Baxter to recognize something, you just hold the object in front of one of its cameras, which are located in the head, in the chest, and at the end of each arm. To program an action, you can move one of Baxter’s two giant arms through the desired motion and select from a number of preprogrammed actions using a pair of dial controls found in each forearm. When you grab one of Baxter’s arms, it feels light as a feather. Its motors compensate in response to your touch, making the heavy limb easy to move through the air.
Sophisticated computer vision software means that even if a plastic widget has toppled over, Baxter will still recognize it. “The shipping and receiving folks have no trouble programming it,” says Vanguard Plastics’ Budnick. “We had it taking parts from one conveyor and putting them onto a table.”
Vanguard Plastics already has 24 conventional robots. They can pick and place objects with fantastic accuracy and speed. But it takes up to a day to program one, and if anything is misaligned they will grab at thin air all day. These robots must also be separated from human workers by protective fencing.
Baxter requires no such barriers. Brooks likes to demonstrate his robot’s safety features by putting his head in the way of one of its giant arms as it swings through a task, smiling as he receives nothing more than a gentle bump. Baxter moves too slowly and gently to do harm, and an array of sonar sensors positioned around its head can detect human movement. The robot also reacts to the sudden change in force that occurs with an unexpected impact and responds by stopping instantly.
The International Federation of Robotics estimates that there are now 1.1 million working robots around the world. About 80 percent of all the work involved in manufacturing a car is now done by machines. In some industries, however, the volume is too low to make automation worthwhile, or the product line changes too rapidly in response to new demand or innovation. This includes small-scale manufacturing but also relatively advanced industries such as aerospace and cell-phone manufacturing.
“It takes significant resources for a company to set up a work environment for a robot,” says Julie Shah, an assistant professor of aeronautics and astronautics at MIT, who studies the role of robots in manufacturing. “It requires suppliers providing materials in a certain way, it requires designing the whole factory infrastructure, and it requires caging the robot. If you need to reprogram these robots, it often takes special expertise or external consultants.”
Some fear that overcoming those obstacles could cost human jobs. But Brooks doesn’t agree. He says Baxter is designed to make human workers more efficient, not to replace them. “An electric drill makes a home contractor more productive,” he says. “Should we ban electric drills so there are more jobs for home contractors? You ask any home contractor that.”
Other experts agree that robots like Baxter could improve U.S. employment prospects in the long term. Willy Shih, a former IBM and Kodak executive who studies the relationship between manufacturing and innovation as a professor of management practice at Harvard Business School, says manufacturing work moved to China partly because it was easier to use humans than to make automation more flexible. Baxter could change that calculation. “To the extent that they can reduce the setup overhead for changing over a line or something, then that becomes really interesting,” he says. “Anything that improves flexibility, that’s a huge deal.”
Boosting the efficiency of small manufacturers could help fuel the economy and drive job creation. By shifting the balance away from low-cost human labor, Baxter and similar robots could perhaps also make manufacturers in the United States and Europe more competitive with their counterparts in Asia. “If you can develop a robot that’s capable of integrating into the human workflow, into the human part of the factory—if it has just a little bit of decision-making ability, a little bit of flexibility—that opens up a new type of manufacturing process, which is what’s really exciting to me,” says Shah.
Back at Vanguard Plastics, Baxter is viewed as just another tool. “There wasn’t any fear, like ‘Oh, I’m gonna be fired’—no, it has nothing to do with that,” says Budnick. He adds that the company plans to buy its own Baxter if the cost can be worked into the budget for the coming year (the robot will cost $22,000). The plan is to have the robot work alongside a person by feeding freshly molded plastic cups into a bagging machine. “The normal robot would populate the conveyor with stacks of 20 cups, and then Baxter would take the 20 and feed a bagger,” Budnick says. “Then all our operator would have to do is take the finished bag and pack them out.”
That would be a humble beginning for a robot as clever as Baxter. But like any other worker, it will start with jobs its coworkers don’t want to do. “I don’t know exactly where it’s going to go,” says Brooks. “But I know it’s going to be a revolution.”
Rodney Brooks is speaking at Technology Review’s annual EmTech MIT conference this October, and Baxter will be joining him at the event.
Register today for EmTech 2012, October 24–25, in Cambridge, Massachusetts.
Become an MIT Technology Review Insider for in-depth analysis and unparalleled perspective.Subscribe today