My favorite new vocabulary word for the year is "co-working," when applied to robots. In June I spoke with Dr. Oskar Von Stryk from the University of Darmstadt, who explained the category to me. Little did we know, Rethink Robotics was just months away from introducing "Baxter," a machine which epitomizes the co-working robot — and which will likely rule the category for the foreseeable future.
Baxter won't get bored
A co-working robot is an industrial robot designed for smaller businesses, and less rigorous tasks. Instead of welding car doors, co-working robots do simple manipulation of objects — like picking them up, and then putting them somewhere else. Imagine multiple buckets of specific Lego pieces, which need to be picked and placed into a single kit of diverse Lego pieces, or a mailer which needs four different pieces of paper inside of it. As someone who has spent thousands of minimum wage hours doing simple sorting tasks (specifically the latter example, and also potato sorting), trying to keep my mind from wandering, I'm glad the robots are finally taking over. Baxter won't get bored, and will turn humans into "operators" instead of simple cogs on an assembly line.
Co-working robots are also need to be easy to program, by definition. An industrial robot might spend an entire year (for instance, a model year of a car) doing the same task repeatedly, which makes bespoke programming worthwhile. In a small business, with smaller runs of products, and more generalized duties for employees, a robot might need to do a different thing each day. Baxter should be up to speed on a new chore in under 30 minutes.
Safety is a "new" concern for robots. Industrial robots work inside a fence, but a co-working robot might work side-by-side with a human. Baxter is flush with sensors to identify potential human obstacles and interference. But safety isn't just limited to human-robot interaction. Von Stryk pointed out to me that there are many simple tasks that take humans a long time, due to workplace safety best-practices. For instance, working with a drill press, a human has to turn the machine entirely off every time he puts an object inside it, then turn the machine back on to press it, then switch the power off to remove, etc. — otherwise a lapse of attention could risk a limb. A co-working robot, while valuable, is not irreplaceable, which makes edgier use of heavy machinery possible without endangering any human hand.
Most important to this category's allure, to my mind at least, is price. Rethink Robotics is offering Baxter for $22,000, which is a serious bargain for a two-armed robot of its capabilities. What's even more exciting is Baxter's modular wrist, which will allow its simple manipulators to be upgraded to hands in the future, apparently by either Rethink or a third party.
A serious bargain for a two-armed robot of its capabilities
What's truly wild is that $22,000 isn't the basement of these co-working prices. I've heard a $10,000 target figure floated, which makes sense for a robot with a bit less than half of Baxter's capabilities: one arm, shorter reach, simpler tasks.
Rethink has the advantage of a head start in the market, and the apparent technical ability to ship a robot with never-before-seen capabilities — "teachability" for a robot is a huge accomplishment in the field. But there are a large number of co-working startups on the horizon, which could make this an exciting, wildly competitive industry in just a few years.
Rethink began as Heartland Robotics, which has some serious financial backing, and was rumored to be bringing a co-working bot to the Chicago's Automate show in January; that's the bot that ultimately came to market as Baxter. Other companies with co-working robots in the works include Redwood Robotics, and Dr. Von Stryk's own BioRob — Stryk's involvement is in making an articulated, safe, cheap hand for one of these robots, allowing it to take on more human-like tasks.
The Wall Street Journal reports that Yaskawa Electric, Universal Robots, Fanuc, and ABB are all working on robots similar to Baxter as well, or at least smaller, "nimbler" versions of their industrial models. I also have to believe iRobot, the defacto leader in military and home robotics, can't be ignoring the co-working space.
But Baxter and his forthcoming friends have meaning beyond just menial manufacturing and collation jobs for smaller businesses. Every quality of these new robots is a leap in the state-of-the-art, and every aspect that these robots will compete against each other on will push that state-of-the-art further. What's thrilling is that required innovation shows a clear path toward robots in the home.
Industrial robots have trended toward rigid joints, increased precision, increased speed, increased power, and increased specialization. Meanwhile, a robot that's safe, easily taught, and cheap, is the exact kind of robot that will eventually be useful in the home.
The robotics industry has been waiting for its "Homebrew Computer Club" moment, the robo-equivalent to when hobbyists could start building the computers they wanted, instead of just tinkering with corporate mainframes after hours.
The robotics industry has been waiting for its "Homebrew Computer Club" moment
In 2003, Pete Markiewicz, my favorite robo-futurist, argued that the homebrew moment for robotics, the equivalent of the integrated circuit that sparked the PC revolution — a commercially viable product, that can be repurposed by hobbyists in less-commercial pursuits — would be a "reliable, environmentally sensitive" robotic body. Well, that moment has arrived: Baxter is the robot body we've been waiting for, a robot that's safe for humans, and yet capable enough to do serious work. Rethink could be the Apple of robotics — in The New York Times, Tony Fadell is quoted saying that Baxter could be the "Macintosh moment for the robot world" — but an "Intel moment" or "Texas Instruments moment" would be just as thrilling, in my opinion.
Baxter is also built on top of the Willow Garage-developed ROS. It's an important moment for the Linux-based operating system, which is already hugely popular with hobbyists and researchers. Baxter represents the big commercial debut for ROS, and shows a path for ROS to become the "Android" of robots — an operating system with heavy corporate backing and development (Willow Garage is Google), but with the flexibility for manufacturers (Rethink is Samsung) of an open source, modifiable code base. Maybe 10 years from now we'll be complaining about Samsung or HTC's heavy ROS skinning. "BotSense," "BotWiz".
I was told that Willow Garage only got a last-minute heads up about Baxter from Rethink — a sign of how competitive and secretive this space has already become, but also a sign of how much freedom and maturity Willow Garage has built into ROS.
For now, the usage of ROS means Baxter is prime for tinkering. In fact, Willow Garage's standard platform for ROS development, the PR-2, offers similar mechanical dexterity to Baxter for roughly 10 times the price.
The feedback loop is delightful
Microsoft's Kinect sensor has been epochal for robotics research, because it put a specialized, obscure capability — infrared depth perception — into a mass market, consumer device. In a similar, slower way, co-working robots will make better hobbyist hardware possible by shipping mechanical hands and arms at volume. And the feedback loop is delightful: the countless researchers building robotic hands right now in universities finally have a market for spinoff companies and a pressure to turn their ideas into products.
As the past decades of over-promise and start-stop in the robotics industry have proven, there is no sure thing in personal robotics. Even Rethink's Baxter could fall flat — it's promising a lot, and a $22,000 under-deliverer won't bode well for the whole co-worker category.
Still, it's not much of a leap to imagine a robot that can identify, pick up, sort, and place objects in a factory being ready, in a few years, to do the same thing with your child's toys and your husband's laundry.