Automaton

...and dirty dishes, house cleaning, and other domestic chores. That's the goal of Anybots, a Silicon Valley startup founded by Trevor Blackwell. The company has been in the news before, but the whole thing is so intriguing we dispatched Automaton correspondent Anders Frick to get more details on the technology. Here's his report:

Economists like to say that the one kind of work you can’t move offshore is personal service, but what if remote-controlled robots become practical?

Trevor Blackwell loves robots, the humanoid kind that populate old sci-fi movies, and like many other roboticists, he thinks there may be a role for them to play around the house. He differs from most, however, in the economic rationale he offers.

Blackwell sees a future in which a low-paid worker from India might remotely control a robot in your kitchen, taking on tasks that today might be assigned to a servant. Blackwell believes that this is the Next Big Thing, and that thousands of homes will be using his robots to clean, cook, and serve meals. This scheme would effectively allow rich countries to import labor -- without the laborer.

To realize that vision, Blackwell founded Anybots in Mountain View, Calif., in 2001, after his last company, Viaweb, was bought by Yahoo for US $45 million in 1998. Blackwell is also a partner in the startup funding firm Y Combinator, which has invested in nearly 60 different startups during the last three years.

He is currently testing both a legged robot, named Dexter, and a wheeled one, named Monty. They now perform only a few, limited tasks, such as serving coffee and operating a hammer drill. It turns out Monty’s the nimbler of the two. “Robots with wheels are both faster and more stable,” Blackwell says.

Each robot has a built-in gyroscope in the torso, position- and force-sensors in the joints and fingers, and magnetic motion sensors in the arms. Their moving parts are actuated by pneumatic plungers and valves, powered by electricity from carbon aerogel ultra capacitors that can go half an hour on a charge.

The 16 cameras carried on different parts of the robots’ bodies supply video to 10 remotely placed monitors. In the beginning, Blackwell says, engineers and technicians will use the robots to steer in particularly dangerous environments -- say, the site of a nuclear or chemical accident. Such work should get the kinks out. That way, when robots go into mass-production for the consumer market, they will be sufficiently reliable, and perhaps also toxic waste-proof, which might come in handy when dealing with some people's dirty socks.

On July 26th, the MIT Museum here in Cambridge, Mass was full of some of the best and brightest roboticists in the area. The Boston chapter of TiE partnered with Robotics Trends to bring together experts to talk about the robotics industry and where it was headed.

Neena Buck, an industry analyst at MIT, and Dan Kara, president of Robotics Trends, introduced the robotics industry to the audience of mostly software entrepreneurs. Helen Greiner, co-founder and chairman of iRobot, gave a keynote about her company and the lessons learned over the last fifteen years. Finally, a panel spent some time answering questions from moderator Dan Kara and the audience. The panel was comprised of a Media Lab PhD candidate named Cory Kidd, also the founder of company Intuitive Automata; Joe Jones, CTO and Co-Founder of Q Robotics and also one of the inventors of the Roomba; Rory MacKean, R&D Manager at Mobile Robots (formerly ActivMedia); and Chris Wallsmith, CKO at Bluefin Robotics (he also has the dubious honor of being my boss).

The panel was fascinating, not just in terms of the answers they gave to the questions, but also to see what sorts of questions were asked by the not-necessarily-roboticist audience. A few interesting points and observations:

• Asia vs the US: there's a well known split in the attitudes toward robots in the US versus in Japan and South Korea. In Asia, it goes, robots are often humanoid (or canine-oid, in the case of Aibo), are meant to interact directly with people, and are thought of --and designed to be -- as pets or companions. In the US, robots are for "dull/dirty/dangerous" tasks like manufacturing or defense and are generally thought of as tools. This may be changing in the US, though. Helen Greiner had stories of Roomba customers asking for *their* Roomba to be repaired, not a replacement unit. Military PackBot operators give awards to the robots as though they are part of the human team and, like the Roomba owners, want their own robot repaired, not to have a new one sent to them. It will be interesting to see how these attitudes drive designs of the next generation of US robots, and whether the US and Asia begin to converge on their designs.
• The "killer app": there were many questions from the audience about what the panelists thought the "killer app" was for the robotics industry -- not a surprising question from those who work in software. What was surprising was the panel's almost unanimous response: there is none, because robots will literally be everywhere. Chris Wallsmith pointed out that robots are much like computers; that is, computers are everywhere -- your laptop, your cell phone, your car, your calculator -- but people don't call them computers. Similarly, he said, your car will be robotic, your kitchen will be robotic, your personal fitness trainer will be robotic... but they'll be called cars, kitchens, and trainers. Not robots.
• Training for robotics: a hypothetical investor in the audience asked what one should look for in evaluating the experience of people proposing a new robotic technology to VCs. The panelists all had different answers -- a background in psychology may help with the design of interfaces and interactions; a broad engineering base is needed to build up the electrical, mechanical, and software systems of a robot; membership in the target customer base lends credibility to the design. The only agreement seemed to be that a broadly experienced group is necessary for success.

So where is the industry headed? Everywhere, it seems. The good news is that not a single person in the room seemed at all pessimistic about the robotics industry; there's funding for startups, a healthy US defense research funding source, rapid growth of new technologies and new ways for people to interact with machines, and growing acceptance of robots working for and with humans. It's an exciting time.

Not Leonardo the artist. No, not the ninja turtle, either. Leonardo is a gremlin-like robot at the MIT Media Lab who was the main attraction in a series of user studies a couple of weeks ago, one of which I got to participate in.

Leonardo was built for the Personal Robots group (I seem to recall them being called "Robotic Life" at one point), headed by Dr. Cynthia Breazeal, to study social interaction with robots. Leonardo can't walk or talk, but he can make a few facial expressions and manipulate a few objects with his eerily lifelike (though not very dextrous) little hands.

The study I was participating in was part of post-doc Andrea Thomaz's research into how humans understand the learning process and how machines can learn from them. She asked me to see if I could teach Leonardo (and if I recognized when I had taught him) to perform a few tasks on a toy box in front of him: pushing a button to change light colors, flipping two switches, and trying to learn the right combinations to open and close the box. Working with Leonardo was a little strange. I'd seen pictures before, but was surprised to find him almost three feet tall (well taller than me when he was standing on a desk). Interacting with him at first felt awkward, but soon I was learning forward on the desk gently urging him to learn what he needed to learn, much in the same way as I might teach a toddler to tie his shoes. He could only respond to a limited set of commands like "Try to flip the switch right, Leo" and to feedback like "Not quite!" or "Good job, Leo!". In the end, I sent him into an infinite loop of switch-flipping (ah, bugs), so my robot-teaching prowess remains unknown. But it was my first time personally interacting with a "humanoid" robot.

And I just noticed now I've been calling him "he" throughout this whole entry. They must have done a good job anthropomorphizing him for me...

At any rate, the study was less about robots than it was about cognition and learning, but I was thrilled to have the chance to check Leonardo out. His command set is limited to whichever program a Lab researcher happens to have loaded on him for the day, so we won't be seeing Leos in our homes pushing buttons and flipping switches for us any time soon. But it was a fascinating glimpse into how I might some day interact with a robot in my house. Hopefully the infinite loops will have been fixed by then.

A report from our correspondent Sally Adee, who earlier this month was at DARPATech, the Pentagon's R&D extravaganza, where she met, among other creatures, a little robotic dog called, well, Little Dog:

This little guy was all the rage at the DARPATech 2007 exhibit hall.

“Can I pet him?” a girl asked the researchers standing next to it. “No,” they told her. “Can I have him?” she insisted. “No, sorry." She thought about it for a minute. “Can I talk to him?” she finally asked. “Well, you could,” the researchers said testily, “but it can't hear you.”

If you’re wondering why the researchers appeared so cranky, it was probably because visitor after visitor, fascinated by the little critter’s attempts to negotiate its obstacle course, only wanted to know how the robot was built, how many actuators it had, what sensors and battery it carried. The perpetually exasperated response: "It’s not about the robot! The important thing here is the software."

Little Dog, developed by Boston Dynamics, is part of DARPA's Learning Locomotion program. DARPA selected six teams, each of which received one Little Dog unit. The teams will have to write locomotion software to make the robot traverse an irregular surface with various obstacles.

Alas, it’s a lot harder to get excited about abstract software. The researchers repeatedly tried to redirect our attention to the big screens above Little Dog showing essentially what was going on in his head. The cameras above the display mapped out the terrain in detail and sent a path to the machine’s processor.

This particular version of Little Dog meticulously plans its every move before it takes its first step. It was a peculiar sight: Little Dog standing transfixed at the start of its obstacle course, staring intently at the road ahead, as the screens flashed and changed and check marks were applied to tasks. It's not everyday you can peek into a dog's brain.

And then Little Dog begins to move. One leg tentatively snakes forward and gingerly tests the ground in front of it before the machine puts its full weight on the foot. This software has a lot to keep track of: it has to distribute the machine’s weight properly to keep Little Dog from losing its balance, all the while locating itself within its physical surroundings. Not all Little Dogs will navigate the obstacle course the same way. UPenn’s Little Dog, for instance, has a different locomotion software than MIT’s Little Dog.

Despite its name, Little Dog's shape was a matter of debate -- a Rorschach test of sorts. When observers compared his movements to those of a cat, cat people immediately came out of the woodwork to point out that his movements much more accurately resembled those of a roach. “His legs aren’t articulated at all like a cat’s,” said Spectrum senior editor Jean Kumagai, herself a cat person.

But everyone agreed that the least of all resembled a dog. At which point the beleaguered researchers would again try to steer the conversation back to the abstract code powering the machine.

Keepon, according to its creators, is:

...designed to interact with children by communicating attention and emotion. It has four degrees of freedom: attention is directed by turning +/-180° and nodding +/-40°, while emotion is expressed by rocking side-to-side +/-25° and bobbing up to 15mm. In addition to his role as a BeatBot, Keepon is used in research on social development and communication.

This little guy is capable of reacting to visual stimuli and of totally rocking out. The video below is both a neat demo of Keepon in public and an excellent music video in and of itself.

Another music video here, and a simpler video of Keepon in the lab reacting to objects here.

Although some people claim their mouths operate independently of their brains, that's not usually the case. The brain sends neurological signals to the larynx, which converts them into sound. Now, what if we could use those larynx nerve signals to control things?

That's exactly what a company called Ambient is doing. Its Audeo technology basically converts "unspoken speech" (neurological signals flowing through larynx nerves when a person thinks about speaking) into control commands that can be used to guide a motorized wheelchair (video above) or synthesize speech. Pretty amazing!

The company apparently stole the show this month at National Instrument's NI Week in Austin. Ambient's founder and CEO, Michael Callahan, gave a demonstration of the company's "thought-controlled" wheelchair and "thought-to-speech" translation system. (You can see the demo at the NI Week video page; it's the last segment, called "Algorithm Engineering," on the August 7 list.)

To use the system, a person wears a lightweight sensor band around the neck. The band picks up the larynx nerve signals and transmits them wirelessly to a remote computer (don't worry about "mind wiretapping" -- the transmission is encrypted.) The remote computer uses NI LabVIEW and signal processing algorithms to interpret the nerve-impulse patterns and translate them into the right commands.

The system is not plug-and-play. It does require some training until its algorithms learn to "read your mind" (accuracy is above 70 percent). But at least it doesn't require Matrix-style brain interfaces or a tangle of EEG electrodes wrapped around your head.

Callahan, a graduate student at the University of Illinois at Urbana-Champaign, hopes to commercialize the technology to improve the lives of severely disable people with spinal cord injuries or such neurological disorders as ALS and cerebral palsy. (The company is backed by the Rehabilitation Institute of Chicago.)

OK -- not exactly related to robotics, but very cybernetic nonetheless. I wonder what things we might control with this technology one day. Any guesses?

I finally had time to read Robin Marantz Henig's 8000-word piece on sociable robots in the New York Times Magazine. In the article, Henig, a contributing writer for the magazine, describes what scientists mean when they talk about "sociable robots," how such robots were designed to learn by interacting with their environments, and what are the issues involving robot learning, robot emotion, and robot boyfriends.

Henig does a great job explaining how the robots work, sometimes by "peeking behind the curtain" -- the robots are mostly MIT robots, old and new, including the metal torso Cog, the bushy-eyebrowed Kismet, the talkative head Mertz, the mop-topped Autom, the Gremlinlike Leonardo, the skyscraperish Domo, and the rubbery bulgy-eyed Rodney (OK, joking about this last one).

More interesting, perhaps, Henig describes instances in which the robots misbehave, or work in a somewhat disappointing way, and hey, that's how engineering happens in the real world, so it was neat to see that in the article as well (by the way, I loved the cover headline, which to me captures the essence of this emerging field: "It Understands (Sort Of)." An excerpt:

Today’s humanoids are not the sophisticated machines we might have expected by now, which just shows how complicated a task it was that scientists embarked on 15 years ago when they began working on a robot that could think. . . . They are, instead, hunks of metal tethered to computers, which need their human designers to get them going and to smooth the hiccups along the way.

But these early incarnations of sociable robots are also much more than meets the eye. Bill Gates has said that personal robotics today is at the stage that personal computers were in the mid-1970s. . . . In much the same way, the robots being built today, still unwieldy and temperamental even in the most capable hands, probably offer only hints of the way we might be using robots in another 30 years.

After reading the article, I wanted to see some of those machines in action, and it's just great you can find so many videos of them (the Times posted a bunch on the article's web page). But as a writer myself I also wanted to know more about Henig's experience writing the article. Having just returned from vacation, she was kind enough to promptly answer my questions -- thanks, Robin! (Follow the link below to read the Q&A.)

My company is pretty firmly entrenched in the defense industry. In fact, many robotics companies are -- defense contracting is a good way to pay the bills while growing other areas of research and development. But while robots are really amazing things to work on in and of themselves, the technology is slowly advancing toward greater capability and autonomy -- and for those of us working defense contracts, this has some uncomfortable implications.

Bluefin's AUVs aren't weapons (when people I ask, I remind them that there is already a word for an autonomous submarine that explodes -- "torpedo") and most other companies aren't actively weaponizing their robots. To date the bulk of military robotics has been oriented toward surveillance, security, and disposal of mines and IEDs -- situations where most everyone can agree that it's a good idea to keep a human out of the way.

But things are changing. Even if companies aren't putting on guns, they're at least putting on gun mounts. Early last month Wired reported on the newly weaponized ground robots. Other companies are building in weapons payload options: recently a Reaper aerial drone made history as the first Army unmanned military vehicle to kill (thank you for the correction, Kevin); its remote operators used it to locate two men suspected of placing an IED and dropped its "precision munitions" on the targets.

What do the users of these robots think? At the OceanTech Expo in early September, I attended an AUV panel; one of the panelists, Bill Schopfel, is the event manager at the Office of Naval Research. He spoke specifically to the role of robotic vehicles in underwater mine countermeasures -- he says for the forseeable future, the decision to engage and neutralize mines will not be autonomous; even if the vehicle is capable of performing neutralization measures, there will still be a person in the loop who is making the decision to engage. With respect to autonomous vehicles that operate without a human's control, a DoD proposal from last year discusses the idea that humans target humans and machines target machines -- though that proposal has not yet passed legal review.

The Army's Future Combat Systems initiative is becoming a reality, but it really demands thought and careful consideration of the ways we're deciding to employ technology -- though our military needs demand immediate technological solutions. How can we make sure ethics and technology develop at the same pace?

Paleo-Future, a wonderful blog that offers a "look into the future that never was," unearthed a 1986 article in The Futurist magazine about the future of personal robots. The future as envisioned in 1986, that is.

The Futurist article, which includes a photo of an Omnibot carrying snacks to a smiling couple [right], makes for an entertaining read, saying at one point that "future robots will be complete home-entertainment centers, able to sing and dance and tell jokes..."

Well, unfortunately a future of pervasive snack-fetching, dancing robots roaming around the house is is not quite here yet. It appears that one big obstacle has been making the darned machines autonomous, capable of doing things -- entertaining the kids, loading the dishwasher, or whatever -- all by themselves.

Maybe the way to go is using humans to control the robots? iRobot's new ConnectR "virtual visiting" robot is one example. It's a kind of mobile teleconference device to be remote controlled by that absent parent always on a business trip. Another example are Anybots humanoids, also designed to be operated remotely by a person (in this case an outsourced servant in a faraway country).

Check back in 20 years to see which personal robots made it to the living room.

PS: On a forecasting-the-future related note, Spectrum has its share of brilliant and embarrassing predictions, many of which you can see in the "Write & Wrong" article that appeared in our 40th anniversary special issue.

Thanks, SKM!

For anyone in the Boston, Mass area, you might be interested in this event at MIT tomorrow (Wednesday) night, a session called "Robots: The Next Wave of the Robot Revolution" that will "explore the advancing robot invasion across all of those sectors." There's a panel of speakers from a few robotics companies, networking receptions, and recruiting (I'll be there representing Bluefin). There's a small registration fee, though it's free for students.

The previous DARPA Grand Challenge competition -- a trip through the Nevada desert taken by autonomous vehicles-- took two tries to get right; the first year, not a single vehicle made it across the finish line. The second year was a much better showing -- four vehicles finished -- and winner Stanford University took away the $2 million prize.

This year's DARPA Urban Challenge took the robots out of the desert and into a (simulated) city. Teams had to build vehicles capable of "executing simulated military supply missions while merging into moving traffic, navigating traffic circles, negotiating busy intersections, and avoiding obstacles." Since this was the first year of this style of competition, many people wondered if it would have the same problems as the first year in the desert -- lots of failures and no one completing the course.

We needn't have worried. Of the 11 vehicles that were allowed to enter the final round of the competition, six finished the course -- though only three teams, Carnegie Mellon, Stanford, and Virginia Tech, finished under the 6 hour time limit.

The MIT vehicle waits at an intersection as a (human-driven) car makes a turn.
Photo Credit: JOHN VOELCKER

So what drives these vehicles (since it's not humans)? The short answer: lots of sensors and lots of computing power. Nearly all the vehicles had some sort of array of laser range scanners arranged on the front -- though while MIT used more than 10, the UPenn entry got away with just 2. A key player in that technology was Velodyne, developers of a high-def LIDAR unit based on their work in the first two DARPA Challenges -- they stayed out of this year's event in order to continue developing their LIDAR technology. Additionally, LIDAR units designed by IBEO and SICK (an old favorite of DARPA teams) were other popular additions to the sensor suite. Stereo vision complimented the laser sensors, and of course, differential GPS receivers and inertial measurement units (IMUs) were must-haves.

While hardware integration is no easy task, software was just as daunting. A layer of hardware interface ("What does the LIDAR say?") under a layer of navigation and control ("Where am I, where do I have to go, how far do I turn the steering wheel, and how fast do I have to go?) under a layer of behavior ("Hm, a stopped car. Wait behind it, or drive around it?") makes for some intense coding. Take the Carnegie Mellon vehicle, which required over 300,000 lines of code to run the 2007 vehicle. Some COTS tools made this easier for teams such as Virginia Tech, who used LabView to "provide the major functions of the vehicle including image acquisition and processing, systems communication, vehicle health monitoring, and vehicle control. A NI Compact RIO system [provided] steering, throttle, and braking control, as well as reading CAN-bus sensors," said NI representative Trisha McDonell.

With the impressive success of the vehicles on Saturday, is my human-driven car suddenly old fashioned? Not so, say the experts. Forbes had a nice article on the competition, and specifically quoted Stanford team leader Sebastian Thrun:

In the eyes of Stanford's team leader, Sebastian Thrun ... the world is still years away from driverless autos. "I'm positively enthused that this race has a winner," he said. "But we’re witnessing the painful birth of a new technology, and this is the first of many hours of labor."

Fair enough, Dr. Thrun. I'll settle for a car that can park itself for the time being.

Special thanks to John Voelcker for insight and photos from the field

The United States Department of Defense