Automaton iconAutomaton

Hexapod Robot Plays Beethoven

chiara piano playing robot

Just like humans, every robot comes with its own unique musical stylin'. Well, mostly unique. Except for the ones that are made from all the same parts with identical programming. But this hexapod, Chiara, has certainly found a comfy little niche for itself in the robotic classical piano world, by plonking away at some Beethoven. I love how this video takes us through the thought process of the robot (or whatever you want to call it), from raw vision to blob detection to kinematics. Have a listen:

Chiara itself is an open source educational robot developed by Carnegie Mellon University. It runs a free programming language called Tekkotsu, and this particular demo was put together by Ashwin Iyengar, a high school student. Nicely done, Ashwin, and good choice of music.

[ Chiara ]

[ Tekkotsu ]

You (YOU!) Can Take Stanford's 'Intro to AI' Course Next Quarter, For Free

stanford artificial intelligence ai course

Stanford has been offering portions of its robotics coursework online for a few years now, but professors Sebastian Thrun and Peter Norvig are kicking things up a notch (okay, lots of notches) with next semester's CS221: Introduction to Artificial Intelligence. For the first time, you can take this course, along with several hundred Stanford undergrads, without having to fill out an application, pay tuition, or live in a dorm.

This is more than just downloading materials and following along with a live stream; you're actually going to have to do all the same work as the Stanford students. There's a book you'll need to get. There will be at least 10 hours per week of studying, along with weekly graded homework assignments. The professors will be available to answer your questions. You can look forward to a midterm exam and final exam. If you survive, you'll get a certificate of completion from the instructors, along with a final grade that you can compare to the grades of all those supersmart kids at Stanford.

You won't technically earn credits for the course unless you're a Stanford student, but for all practical purposes, you'll be getting the exact same knowledge and experience -- transmitted directly to you by none other than two living Jedis of modern AI. Thrun, director of the Stanford AI Lab, led the team that won the 2005 DARPA Grand Challenge, and, more recently, he helped develop the Google self-driving car. Norvig, a former scientist at Sun and NASA, is now director of research at Google and co-author of the leading textbook on AI.

Here's how it will all work: Anyone can sign up for the course online. It starts on October 2nd and lasts 10 weeks. Each 75 minute lecture (there are two per week) gets videotaped and chopped up into 15 minute chunks that you can stream whenever you want, and homework, quizzes, and exams are all digitized and completed over the internet. Professor Thrun gave us a few more details:

Grading will be automated. But we are recording video specifically to help students who got the answers wrong. We will use the exact same questions for everyone, including the Stanford students. In this way we can actually compare how well everyone is doing.

We will use something akin to Google Moderator to make sure Peter and I answer the most pressing questions. Our hypothesis is that even in a class of 10,000, there will only be a fixed number of really interesting questions (like 15 per week). There exist tools to find them.

As of yesterday, which is only the third day that the course has been available, over 10,000 students are already signed up, and since enrollment is open until September 10th, it's entirely possible that a couple hundred thousand people could end up taking this course. Sounds daunting, but professor Thrun is optimistic about the whole thing:

I am very excited. Teaching many students online has always been my dream. This quarter I get to affect more students than in my entire career before. And yes, we are already beyond my expectations, just 3 days in.

You can sign up for yourself at the link below, and keep up to date through the class Twitter feed here.

[ Introduction to Artificial Intelligence ]

RIBA II Healthcare Robot Gets Bigger Muscles, Cuter Ears

We first met RIBA (or, RIBA-I as we should start calling it) back in 2009, although the assistive robot has been around since 2004. Developed in a partnership between RIKEN (a natural sciences research institute in Japan) and Tokai Rubber Industries, RIBA's job is to lift people when asked nicely. Seems trivial, yes, but if you don't have a robot bear around to help you out, you either need to use an awkward and uncomfortable piece of machinery or have a person do it, and if you haven't noticed, we humans are hefty and getting heftier. This is especially problematic for healthcare workers who have to lift patients frequently, and often get injured doing so.

RIBA is snuggly soft to make lifting comfortable and fun, and the pronounced bearishness is there to help patients relax. This new and improved robot (RIBA-II) features springs to help it lift more weight, and it responds to both touch and voice commands:

So let's just run the numbers real quick on RIBA's increasing buffness: in 2006 RIBA could lift 40 pounds, in 2009 RIBA could lift 135 pounds, and RIBA II is now up to 175 pounds. Let's see, that would mean that RIBA will be cracking the 1,000 pound mark by around 2040, which might just be enough to keep up with the pace with which we humans are packing on weight. But it's gonna be close.

[ Press Release ] via [ Reuters ]

PR2 Successfully Bakes Giant Cookie From Scratch

This is it, folks. The epitome of robotics. After some practice runs, PR2 has successfully managed to bake itself a cookie completely from scratch:

Not being a baker, I'm not sure if it's normal for the cookie to look more or less the same coming out of the oven as it does going in. But whatever, it's got chocolate and sugar and butter in it, and we can just act all snooty and say that the cookie has been "deconstructed" by the robot in a spectacular show of culinary skill.

Obviously, there's still a bit of optimizing to be done with BakeBot here, and I'm sure that the students at MIT CSAIL are already putting in lots of overtime running this routine over and over again to try out new algorithms (and recipes). We can all be thankful that they're making this delicious sacrifice in a noble effort to extend the baking capabilities of robots everywhere while keeping their chocolate cravings at bay. Robotics sure is tough, isn't it?


Foxconn To Replace Human Workers With One Million Robots

Foxconn, an electronics manufacturer from Taiwan with huge factories in China, generates about 40 percent of the global consumer electronics revenue by creating things like iPhones and computer components on giant assembly lines staffed by humans. Until recently, you'd probably never heard of Foxconn, but a series of worker suicides made us all take a hard look at where our electronics were coming from. Foxconn has made some improvements (including nets around tall buildings), but by all accounts, the core of the problem (the work) remains "repetitive, exhausting, and alienating."

Yesterday, Foxconn announced (at an employee dance party of all places) that they're planning on buying some robots to replace their human workforce. And by some robots, they mean one million robots over the next three years. So for every one robot Foxconn currently has working at their manufacturing plants, they're going to buy a hundred more.

At this point, it's not sounding like Foxconn is trying to augment its human workforce with robots to make things easier on the humans. Foxconn employs something like 1.2 million people, and it's not too much of a stretch to imagine that one robot could probably work as efficiently as 1.2 humans, especially considering that the robot can be less productive (even substantially less productive) if it just works more hours than a single human is capable of. I'm not suggesting that Foxconn is considering replacing the entirety of its production line -- which by the way will keep expanding at a furious pace -- with robots, but when you think about how much they spend providing food and housing for their human workers as well as the recent suicides, you can sort of see where their train of thought is heading here: This could be a shift from "mostly human" to "mostly robot," with about a million jobs in the balance.

While Foxconn's manufacturing plants are certainly not ideal places for humans to work, lots of people do currently work there, and those Foxconn employees depend on their jobs to the same extent that the rest of us do. I think we all realize that robots replacing humans when it comes to repetitive manufacturing jobs is a gradual inevitability, but it's a bit of a shock to consider a million robots over such a short span of time.

Rumor has it (and we should stress that these are rumors) that the actual robots being deployed at the Foxconn plants will come from ABB. Specifically, they'll be ABB's Frida robot, although funnily enough, ABB "insists that its robot isn't designed to replace human workers, but rather to work alongside them:"

So, in a nutshell, this might be great news for ABB. It might be good news for Foxconn. But for any of the million or so people with a job, a home, and a life at a Foxconn plant, things may be about to get even worse.

[ Xinhua ] via [ Hizook ]

Sarcos Exoskeleton Bringing Iron Man Suit Closer To Reality

sarcos exoskeleton robot suit

Ever wanted to become Iron Man? Here's some good news: Sarcos recently said that its second-generation exoskeleton robot suit, XOS 2, is now five years away from production. IEEE Spectrum contributor Susan Karlin writes:

The wearable robotics suit augments the operator's strength by using a system of high-pressure hydraulics, sensors, actuators, and controllers to bear the weight of an object, while leaving its wearer agile enough to kick a soccer ball. It's also lighter, stronger, and more environmentally resistant, and it uses half the power of the company's first exoskeleton, XOS 1, which rolled out in 2008. The XOS 2 has been nicknamed the Iron Man suit in homage to the high-tech power suit in the comics and movies.

We first wrote about the Sarcos exoskeleton more than five years ago, when it was just a prototype developed as part of a DARPA program. Since then, Sarcos, now a division of U.S. defense contractor Raytheon, has significantly improved the device. The XOS 2 exoskeleton is designed to lighten a soldier's load and help the military reduce injuries. It also lets you pretend you're Tony Stark:

Several other companies and universities are developing exoskeletons to help not just soldiers but also the elderly and other people who might need assistance to walk, climb stairs, and carry things around.

Most notably, Japanese firm Cyberdyne has created a robot suit called Hybrid Assistive Limb, or HAL, which is commercially available in Japan. Automaton writer Evan Ackerman tested the HAL system at CES in January, becoming the first man in the United States to try out the device.

Then there's U.C. Berkelely spinoff Berkeley Bionics, which last year introduced its eLEGS robotic exoskeleton, a very impressive system that is helping paraplegics to stand up and walk. The company is currently testing eLEGS with a select group of rehab centers, hoping to make it available for purchase in the next year or so.

Image and video: Raytheon

This Just In: Robots Prefer Tecate Over Bud Light

Presented without comment.

Well, mostly without comment. I will briefly mention that PR2 also spurns Bud Light, suggesting that robots do in fact have an inherent bias against beers that are terrible. And as far as Darwin goes, my guess is that he'd probably be in favor of Red Stripe but against Pabst Blue Ribbon. Nice job, Darwin: next time I throw a party, you're definitely invited.

If you want your own Darwin-OP to party with, by the way, he can be yours for a mere $12,000. Rehab sold separately.

[ Trossen Robotics ]

WANT: Volkswagen Demonstrates Production-Level Automotive Autopilot on Video

Volkswagen announced their Temporary Autopilot (TAP) system last month, and it's just shown up on video. If anything, it works better than advertised, and includes some innovative features that do their best to keep you safe, even if you completely zone out:

As you can see, this TAP system has been integrated into a production car, and uses production-level radar, camera, and ultrasonic sensors along with by a laser scanner and an electronic horizon to do everything that it does. In other words, there's no crazy custom electronics involved, and nothing that could keep a system like this from becoming (say) an optional extra in a production car relatively soon.

A big stumbling block for this kind of thing is the issue of liability and who is (or isn't) in control of the car, and Volkswagen very deliberately includes the following in their press release:

"The driver always retains driving responsibility and is always in control. The driver can override or deactivate the system at any time and must continually monitor it.” TAP always offers the driver an optimal degree of automation as a function of the driving situation, acquisition of the surroundings and driver and system states. It is intended to prevent accidents due to driving errors by an inattentive, distracted driver. ...Drivers must still continually focus their attention on the road, so that they can intervene in safety-critical situations at any time.

In other words, this is not (not) a substitute for a human driver. It's not even really an autonomous system, in the strictest sense. It's there in case you (the human) fail at driving for whatever reason, but it's not designed to enable you to not pay attention to the road. In fact, as the video shows, the TAP forces you to pay attention using audio alerts and what looks to be a rather aggressive tap on the brakes if it thinks you're ignoring it. Overall, this is a big step, but still just a step, towards the eventual goal of complete automotive autonomy.

More steps please.

[ HAVEit ] via [ ]

Darwin-OP Learns To Play Dance Dance Revolution

If you're either too old or not old enough to remember the heyday of Dance Dance Revolution (aka DDR), that's totally fine. You're not missing much. It was (is?) a video game that involves "dancing" (I'm actually making air quotes over here) by standing on various combinations of floor sensors as instructed by a video screen in time to music of dubious quality but emphatic volume.

The primary appeal of DDR, as far as I've been able to tell, is watching your friends degenerate into crazy people while playing the game, and unfortunately, robots (even the sweaty ones) can't really offer this same level of entertainment (despite their mad dancing skills). I mean, if I was a robot tasked with playing DDR, I'd probably be wondering what all the fuss was about. You see an arrow, you make the movement, what's the big deal?

For this Purdue University Darwin-OP, it's not a big deal at all. A student there has decided that his summer robotics research project is going to be to teach Darwin to play DDR, which is so far looking to be an entirely possible task, with the help of a slick custom robots-only dance pad:

At the moment, Darwin relies on a balancing bar for stability and to enable faster moves, but you hardcore DDR players should be familiar with the safety bar on the arcade machines that could be used (by crafty humans) for essentially the same purpose. In the works is tuning the robot's vision system to allow it to play DDR for real, and bar-free stability may come after that. Is anyone else thinking that Robot DDR would make a great new RoboGames event? No? Just me? Oh.

Via [ Kotaku ]

Love Is in the Water For Some Reason at RoboSub 2011

It's the 14th year of AUVSI's RoboSub competition, which of course means that all of this year's challenges are love-themed. You know, because of 14. Valentine's Day. It's the 14th. Of February. Yeah, I dunno, if it was me I would have gone for a The Hunt for Red October theme or something a little, uh, edgier.

Anyway, the competition took place from July 12 to 17 at the U.S. Navy's SPAWAR System Center down in San Diego, where nearly 30 teams (including both high school and international teams) unleashed their autonomous robot submarines against a hapless swimming pool filled with gates, buoys, paths to follow, objects to retrieve, and targets to torpedo. If you're wondering why this is so hard, here's a comment on last year's competition from the 2010 Maryland team's advisor:

Some more food for thought on how difficult the competition is: navigation for subs can’t rely on GPS (GPS signals only penetrate a few inches in the water), there’s no contact with the ground (so you can’t use encoders), and substantial random currents render dead reckoning worthless. The teams that can afford it buy a Doppler Velocity Logger (costs around $20k) to give them ground-track velocity.

Water is a difficult environment in which to maneuver. Teams must either make a vehicle that is so large that it can’t spin out of control or make a vehicle that can control its 3D orientation as well as 3D position. There is a strict penalty for large vehicles, so most teams opt for fancy control. It is worthwhile to note that in some years of this competition only a third of the teams could actually drive underwater in a straight line.

Competition objectives include not just visual tasks which are difficult underwater (light patterns caused by sunlight passing through the surface are called caustics and make shape recognition challenging) but acoustic objectives as well. The highest-scoring competition objectives require a passive sonar system.

So yeah, even tasks that would be a dead cinch for a robot driving on land is extremely difficult for a robot under the water. I'd go on about this, but it's more fun to just watch the recap vids from all three days of the competition, so here you go:

Part of the competition required each team to put up a website about their robot, and you'll find links to all of those (with tons more info) at the link below.

[ RoboSub ]

Most Commented Posts


IEEE Spectrum's award-winning robotics blog, featuring news, articles, and videos on robots, humanoids, automation, artificial intelligence, and more.
Contact us:

Erico Guizzo
New York, N.Y.
Senior Writer
Evan Ackerman
Berkeley, Calif.
Jason Falconer
Angelica Lim
Tokyo, Japan

Newsletter Sign Up

Sign up for the Automaton newsletter and get biweekly updates about robotics, automation, and AI, all delivered directly to your inbox.

Load More