Assembly bot Baxter wants to get close to you (Q&A)
Boston startup Rethink Robotics thinks its new humanoid factory worker is so low-cost and easy to set up that it could change U.S. manufacturing. Crave spoke to Rethink founder Rodney Brooks.
If Baxter had a favorite band, it might be The Carpenters. Rethink Robotics' new droid could hum "Close to You" while it gets cozy with human workers along the assembly line.
The Boston-based startup launches Baxter today, billing it as a revolutionary humanoid robot that could help stem the tide of manufacturing going overseas for cheap labor.
For one thing, Baxter itself is surprisingly low-cost. Priced at $22,000 including software upgrades, it goes for far less than traditional industrial robots and puts automation in the hands of small and midsize companies that may not have been able to afford it. Labs and universities are also expected to show interest.
Also, unlike most factory robots, Baxter doesn't require a safety cage. People can work alongside the droid, which is covered in soft materials in case of impact. Sensors tell it when people are near, and it will stop moving if it does make contact with something unexpected.
Meanwhile, computer vision and common sense know-how allows it to do things like adapt to a situation (for instance, if it drops a part it can continue the task by grabbing another).
We spoke with legendary roboticist Rodney Brooks, an iRobot founder and MIT professor emeritus who is Rethink's founder and CTO. Check out his answers and the Baxter photo gallery below.
Q: What's special about Baxter?
The robot has human form, its arms are a little longer than a human's, and it's designed to go into exactly the same workspace as the people around it. It's also designed so that an ordinary factory worker can train it to do a task in minutes. It's applying what IT has done for the office worker or ordinary consumers, giving them powerful tools and smartphones, to industrial automation to increase productivity of regular factory workers.
How about its features?
Each arm has seven joints. We're shipping two attachable hands -- a suction gripper and an electric two-fingered gripper. The head has got a screen on it, some eyes, and facial expressions to tell people what it's about to do. When you want to train the robot, you grab its arm by the wrist and move the arm around. There various buttons on the robot for training, as well as iconic menu options on the screen. It's a platform in the same way that a smartphone is a hardware platform which gets upgraded at a one frequency, and the software gets upgraded at another frequency. That's what our robot will be.
What will it be doing?
Picking stuff off conveyors, putting it in boxes, moving stuff around, putting stuff on test equipment. Later it will be machine tending, putting things into machines, pressing buttons, a lot of testing along production lines. Later it will be more box packing, with more and more capabilities over time. These are dull, boring jobs but every production facility has lots of them.
How is Baxter unique compared with other humanoid industrial robots like Motoman?
Our robot does not require a safety cage. When Baxter is operating, you walk right in with it, you hug it, you grab it. It's much more collaborative. It's got soft rubber surfaces and is constantly monitoring everything around it, so it can detect people from quite a distance off as they come close to it... It will glance over at you as you come up to it, so you know it's seen you. When you touch it, it stops. I routinely walk in at full speed and have Baxter smash me on the head. My engineers know that's going to happen, so they built it to be very safe.
So how can Baxter help American manufacturers?
Here's the key differentiator, apart from the safety, apart from the cost. If you buy a normal industrial robot, it takes weeks to get a quote, you get a systems integrator to put the system together, and it might take four or five months before an arm is delivered to the integrator, you load it up with other software, it's programmed by an engineer, and meanwhile, you have to reorganize your production floor to have safety cages, and put a lot of space around the robot. It's a 9- to 18-month process to get a robot installed.
There are 300,000 small and medium manufacturers in the U.S. who do not have any sort of robot in them. Our goal is it will take two hours from delivery of the robot to when it's doing productive work. In our trials it's been less than an hour. A person who has never seen a robot before can learn to train it in 5 to 10 minutes.
If you think about the transition from mainframes to PCs to smartphones, we're doing that. It's a smart robot, but it's bought and sold and packaged more like a consumer product. It's easy to use. You can take it out of the box and use it.
You co-founded iRobot, which produces military and consumer robots. Why did you want to make industrial robots?
I spent a lot of time in China, Taiwan, and Hong Kong figuring out how to do low-cost manufacturing. I realized it's not a long-term sustainable model. If you look historically, we outsourced manufacturing for low-cost labor to Japan after World War II. The standard of living went up, and we moved to South Korea, then we moved to Taiwan, then we moved to southern China... Labor costs have gone up in China as the standard of living has gone up as people demand a better work environment, and that's the story of Apple and their manufacturers.
So I started thinking of all the advantages of not manufacturing in China. The cost of doing the engineering with people flying to China all the time is immense. We're manufacturing this robot in the U.S., all over the country, and it's a much more fast, innovative way of engineering. And then there's IP protection. Everyone I know who's ever manufactured in China has a story about seeing a version of their product show up somewhere else. And I've certainly had that experience with Roomba, many times, many models.
How do you think this will impact the spread of robots in general?
The Roomba is special-purpose. The software development kit on Baxter will make it general purpose. This robot has two arms, you can do different things with it, and it can do tasks. Small companies we've talked to see this as a way to make them more competitive, so they can go after more contracts and increase their workload, and having their workers do more productive stuff.
When I first came to the U.S. from Australia in 1977, as far as I knew there were a total of three mobile robots in the world, though there may have been some in the Soviet Union that we didn't know about. iRobot has now shipped over 8 million, so that's a pretty good transformation.
