SciAm: The Year in Robots

Just the other day, I was complaining that Scientific American editors left out robotics from their top 25 science stories of 2007. I guess I spoke too soon because today I noticed that SciAm has now published an article focused 100% on robotic achievements in 2007. The article, The Year in Robots, discusses a number of major robot achievements including the Urban Challenge, Toyota's violin playing robot, the slightly updated Honda ASIMO, and NASA's medical robots among many others. The image gallery associated with the article is also a nice touch and definitely makes a visit to the site well worth it.

Nano robots and human sperm cells

Researchers at Cornell University are trying to figure out how to use the same mechanism that powers human sperm cells in medical nano robots. According to a CBC article (my source for this post),

By deconstructing the stages in the biological pathway sperm cells use to generate their relentless energy, researchers at the Cornell University College of Veterinary Medicine in the United States hope to recreate that process in an artificial device.

Medical nano robots are a very promising research direction for creating efficient mechanisms for medication delivery and for directly fighting disease, i.e., nano robots working along with natural white cells to fight viruses inside the human body. Medical nano robots can help us eliminate the problem of antibiotic resistant viruses once and for all and it is certainly one application of robotics that I am mostly looking forward to.

The CBC article, "Sperm power could drive nano-scale robots," has a nice overview of what the researchers have achieved so far and what they hope to do in the future given sufficient funding to continue their work.

Robot chair concept

A robotic chair that follows people around a library was the graduating project for a student at the Design Academy Eindhoven. The video below shows the concept chair activated using an access card and following a person around the library. Towards the end, the video also shows how a group of chairs could automatically be reconfigured to a particular formation upon a single user's command.

As noted earlier, this is not a working system but a concept demonstration. As it is noted on YouTube,

This is a concept. This model has a R/C remote controlled “robot” inside.

Regardless, I think it is a cool idea. I also believe that we already have all the necessary technology to develop and deploy these robotic chairs. The technology developed for the museum tour-guide robots has already been demonstrated to work well in crowded spaces and it could be used to make these concept robot chairs into a real product.

An improved image search algorithm

MIT's Technology Review reports on new research in image-based retrieval that could have significant impact in improving online image search engines such as Google's image search feature. The work comes from the the University of California, San Diego (UCSD,) and more specifically Nuno Vasconcelos' Statistical Visual Computing Laboratory (Professor Vasconcelos is currently the head of this lab.) The researchers have improved upon current machine learning algorithms to allow computers to more accurately classify images and their content using information in the image as opposed to textual information found in the surrounding text or the image filename as it is commonly done by most popular search engines today.

The goals of Vasconcelos' group are very ambitious. As he explains in one of his papers,

One of SVCL’s goals is to develop systems capable of retrieving images because they understand them and are thus able to represent their content in a form intuitive to humans. Drawing strongly on computer-vision and machine-learning research, this effort explores many issues in image representation and intelligent system design including the evaluation of image similarity, the automatic annotation of images with descriptive captions, the ability to understand user feedback during image search, and the design of indexing structures that can be searched efficiently.

In case you are wondering, Google is all over this work as explained in the Technology Review article. So, don't be surprised if eventually this improvements make their way into Google's search engine. And I sure hope this happens some time soon because Google's image search is certainly not very good at the moment. Vasconcelos sure believes that even better algorithms are definitely only a few years away.

Vasconcelos of UCSD suspects that it will be more than five years before computers are able to identify more-difficult concepts, such as happiness, in pictures. But that doesn't mean current research won't be useful before then, he says. "The expectation has to be that [the technology] is more like an aid, not like an answer."

Read: Better, More Accurate Image Search.

Top science stories of 2007: where are the robots?

Scientific American editors and reporters have decided upon published their top 25 science stories of the year. During the last couple of weeks in December, most publications (offline and online) always publish articles about the top stories in any subject; a science publication is no different. Regardless, I was surprised to see that out of 25 stories, the Scientific American editors only included one robot related. The story selected had to do with an effort to test a remote-controlled robot designed for underground mine rescue operations. The location for the tests was the collapsed Crandall Canyon Utah mine which had cost the lives of six mine and three rescue workers. Rescuers earned some valuable experience using a robot in a mine rescue operation and there is hope that this will translate to better rescue robots in the near future.

The rescue robot story was interesting and likely worth a mention. I was surprised, however, that the Urban Challenge robots did not register in SciAm's radar. These robot car autonomously navigated for hours city-like streets obeying the laws of traffic and sharing the road with other vehicles some driven by humans. The Urban Challenge was conquered easily after the first attempt compared to the Grand Challenge that required two takes before any cars could finish the course.

Other than the robots of the Urban Challenge, I think the bomb disposal robots currently used in the Middle East are also worth a mention. These are robots that are saving lives; and this is a fact that cannot be disputed whether one is pro or against the war. The rather bad news are that the military is slowly funding research to make these into weapon carrying machines capable of killing people; it is basically inevitable that autonomous robot soldiers will at some point become a major part of the armed forces of those countries that can afford to have them. Sad, but true!

Read: Top 25 Science Stories of 2007 on SciAm.com.

A great week for iRobot

The most successful robotics company in North America had a great week as 2007 is fast coming to its end. First, iRobot was awarded a $268 million army contract to develop the new generation of bomb disarming robots part of the already successful PackBot line. This contract had originally been awarded to iRobot's rival Robotic FX. It turns out that Robotic FX is a small startup founded by a former iRobot employee. The two companies have been fighting it out in court since last August after iRobot accused Robotic FX of stealing their technology. As an added bonus to the new army contract, the courts also decided in favor of iRobot in their disputes with Robotic FX; this explains why the army canceled the original contract with them and assigned it to iRobot. Of course, this is terrible news for Robotic FX which is now out of business. In addition, its founder has been prohibited from being involved in the robotics industry for 5 years. This is harsh and I sure hope that he has other skills to make a living with.

Interestingly, however, I really expected all the good news to work miracles for iRobot's stock price but alas that didn't really happen. The army contract helped the stock price to gain 4 percent but the court's decision on the patent disputes with Robotic FX hardly influenced it. It could be that the latter happened on Friday and so the market did not have enough time to respond. Next week should be good for iRobot's stock price, I think. In general, iRobot is a good company to own a piece of because let us not forget that its founder is none other than MIT's legendary roboticist Rodney Brooks. And they have demonstrated their ability to design and mass produce useful robot more than any other company in North America.

Japan's 2007 robot of the year

In the 2nd year of the government sponsored Japan's robot of the year award, function over form was the deciding factor. Fanuc's industrial robotic arms were the clear winners over more awe aspiring robots such as Komatsu's fire-fighting remote-controlled robot and Fujitsu's humanoid robot. Fanuc's robots are being used in the food and pharmaceutical industries. If I remember correctly, the winning robot last year was Paro, the therapeutic robot seal. I guess the Japanese judges are making sure that the cutest robot won't always be the winner. After all, the point of this competition is to reward those who design robots to help the country cope with a shortage of labor workers and Fanuc's robots are certainly doing their part.



BBC's article on Japan's robot of the year award: Robotic arms take home top award.

The photo is copyright Katsumi Kasahara/Associated Press.

Robotics jobs wiki

If you are looking for a job in robotics then you should check out IEEE Robotics and Automation Society's (IEEE RAS) Robotics and Automation Open Positions wiki. Here you can find full-time research scientist positions, post-doc and PhD opportunities and finally, tenure track faculty positions. As I am someone who in the not so distant past had to look for work after graduation, I think this wiki is a great resource for job seekers. It is very difficult to find research related positions on the more popular job websites such as Monster.com. The robotics and automation jobs wiki will be a great place to find a good selection of opportunities for those entering the job market after completing their graduate studies. Check out the recent job posting for positions available at NASA's JPL; I wonder if these will involve working on NASA's next generation planetary rovers!

By the way, another great resource for robotics jobs is the robotics-worldwide mailing list. New positions are posted on this mailing list almost daily so it is worth subscribing to it.

VOLTS-IQ SDK version 1.6 released

Vancouver's Braintech has released a new version of their vision SDK for the Microsoft Robotics Studio. The SDK provides routines for object detection and tracking and, in the new version, face detection. This is the best (if not the only one) professionally developed and supported vision SDK available for MSRS.



To download the SDK and for more information, including documentation, demos and the official support forum visit the VOLTS-IQ website here.

Via Rob Sim's (Braintech lead scientist) blog.

3D search engine

A very cool project from the Shape Retrieval and Analysis group at Princeton involves the creation of a 3D-model search engine; queries to the engine can be keywords but also a hand drawn outline of the object to search for. The researchers have constructed a database of 36000 3D models and have developed a JAVA application that allows users to try out the system over the web. The image below shows an example query and the results returned for a 2D outline of a car.


One of the major problems that the team is trying to solve is discovering an efficient representation for 3D shape so that different objects can easily be identified. Some representations can be too general; for example, we can represent a human head and a soccer ball using a sphere but it would be hard to distinguish between the two unless more elements of the shape of each object are taken into account, i.e., the facial characteristics of humans can help differentiate a head from a ball. In addition, a model that is too specific will not be good for recognizing objects in a large database; for example, we would like to discover the features that make a car easy to recognize (a general class of objects) as opposed to a specific model, i.e., an SUV versus a sports car.

The following video provides an overview of how the 3D search engine works and how it can help find items that cannot be easily searched for using keywords only.

The 3D model search engine allows anyone to upload a 3D model of an object in order to help make it better. So, if you have some 3D models lying around, you can (and should) help the Princeton researchers improve their search engine.

AIBO coming back from the dead?

Rumor has it today that Sony will bring back their robot dog AIBO with new features which will include connectivity with a PSP/PS3 and remote control via the PS3's controller. I found out about this via this post on Gamespot so I don't know what the chances of this being real are. The poster says that there is a 53% chance of this being true but how the heck does anyone come up with such a precise probability (this in my opinion is strong evidence that this is a fake)? At any rate, if Sony brings back their robot pet, then at least, they are not making the same mistakes they made the first time around, i.e., the Sony AIBO PS will have a remote control this time around. Hopefully, it will also be cheaper than the previous model.

Pleo review: good or bad news for Ugobe?

PC Magazine editor Lance Ulanoff recently reviewed Pleo, the new robot toy from Ugobe that has been marketed by the company as the first artificial life form. Ugobe finally started delivering Pleo to customers after many months of delays and just in time for the holiday season. The robotic dinosaur toy is priced at $350. I have previously said that in my opinion Ugobe promised too much with their robot toy dinosaur and their failure to deliver will alienate future customers. Lance's (and other reviews such as the one recently published by Engadget) support my opinion even if these reviews seem positive at first. Pleo is not the holly grail of robotics promised to us and it is not likely to be one any time soon.

Why do I think that Lance's review is actually negative and not positive as his first paragraph seems to indicate? Let us take a look at the pros and cons of Pleo from his perspective.

Pros:

• Pleo is cute and charming.


• On the technical side of things, its software is upgradeable via a USB port and an SD card slot.


• The robot is loaded with motors and sensors so one would expect it to sense its environment and act in a natural way; the reviewer suggests that the robot looks intelligent in its behavior (although I believe that his definition of intelligence is very liberal.)


• Pleo's skin is well designed and it feels nice when touched.

Cons:

• Limited appeal after the first few minutes; Pleo grows “mature” in less than one hour of being turned on.


• Life OS is not shipped with the current version so Pleo does little to no learning; even though its software is upgradeable, no upgrades are currently available an none are planned for a few months. But Ugobe has promised to release Life OS and software upgrades some time next year.


• Pleo's batteries have a long recharging time of 3 hours; when fully charged, the batteries can provide only 1 hour of play time!


• It has noisy motors; one may be able to deal with the noise for a while but in the long term, I bet the noise will be a nuisance most parents could do without.


• It is slow moving reminding us all of why AIBO was so boring to play with.


• Pleo is targeted to kids 8-12 years old; as the author says, parents who have to make the choice between a Pleo and an Xbox 360 (or a Nintendo Wii) are not likely to buy a Pleo. Would you?

I don't know about you, but reading the review, even though Lance does not say it directly, one can only conclude that Pleo is not really a great gift for this holiday season but also not even a good robot to boot. Lance suggests in his opening paragraph that Pleo is great but this is no surprise to me considering that at the very beginning of the article PC Magazine has a link for purchasing Pleo earning the publication a hefty commission. It is to their best interest to write a positive review or at least a negative review that to the average reader would pass as positive.

If you want to spend $350 on a robot this holiday season then your money is better spend on buying a Roomba vacuum cleaning robot and a WowWee RoboPanda, i.e., buy two robots worth the money you would waste on a Pleo.

Violin playing robot: Why is it important?

You may have already heard about this but a couple of days ago, Toyota showcased a humanoid robot capable of playing the violin. The video below shows the 152cm-tall robot playing Edward Elgar's "Pomp and Circumstance."

Now, I'm certain that you are wondering what is the usefulness of a violin playing robot? Toyota and many other technology companies in Japan are in a race to develop affordable helper robots to assist an ageing population on daily tasks. The humanoid robots revealed by Toyota are part of a roadmap to start mass producing such machines no more than 5 years from now. But, do elderly people really want a robot that plays the violin or do they actually need a robot that can help them pick up their groceries and do their dishes?

The reason Toyota has developed a violin playing robot is not because the robot's ultimate purpose is to entertain people with its musical talents. Toyota is simply demonstrating the advances in their robotics technology. More specifically, they are showing us how dexterous their robot is and as they said, they plan to make it even better in the next couple of years; I would not be surprised if they achieved such a goal considering that this robot is the product of Toyota's robotics group established no more than 3 years ago. Playing the violin is not an easy task. The fact that this humanoid can do it is impressive to say the least.

Grand Challenges payoff for CMU

It appears that the two successful entries in DARPA's Grand and Urban Challenge events have earned CMU another $14.4 million in funding for research in autonomous Unmanned Ground Vehicles (UGV.) Granted, this is not the first time that CMU has received funding from the military to develop teleoperated and/or fully autonomous UGVs, but the timing of the announcement for this round of funding does make me believe that if anything the final amount must have had something to do with CMU's flawless performance at the Urban Challenge finals last October. That said, the project being funded is well described in CMU's Press Release,

The US Army Tank-Automotive Research, Development and Engineering Center (TARDEC) awarded a $14.4 million contract to develop an advanced, autonomous, unmanned ground vehicle (UGV) to Carnegie Mellon University’s National Robotics Engineering Center (NREC).

The TARDEC contract funds NREC to build an updated version of the center’s successful "Crusher" UGV as part of the Autonomous Platform Demonstrator (APD) program. Additionally, TARDEC plans to add to the contract, on a work directive basis, additional effort for NREC Engineers to develop an Unmanned Ground Vehicle (UGV) end-to-end control architecture and demonstrate the viability of autonomous UGV operations in a relevant environment as part of the Robotic Vehicle Control Architecture (RVCA) program.

Interestingly, a quick look at the projects versus capabilities page on the NREC website (click here) reveals that 3 previous projects nicely complement each other in order to fulfill the goals of this new project. Specifically, the projects that I am talking about are the Crusher, PerceptOR and Urban Challenge.

Finally, and for no other reason other than the fact that Crusher is one cool vehicle, enjoy the following video showing off Crusher's ability to effortlessly navigate very harsh terrain.

The uBot robot for mobile manipulation

The uBot series of robots from University of Massachusetts at Amherst robotic groups is a well designed dynamically stable mobile manipulation platform. The uBot-5 shown in the photo to the left is the latest model of a long series of robots developed at the Laboratory for Perceptual Robotics since 1997 with the uBot-0.5. After 10 years of development the uBots have evolved from small reactive heat-seeking robots to very skilled, dynamically stable mobile manipulators. The uBot is a wheeled robot that is dynamically stable similar to a Segway Robotic Mobility Platform (RMP.)

The uBot-5 is a small and lightweight research platform for mobile manipulation. It was designed to be an economical robot that is highly capable, durable, and safe to operate. When equipped with a LCD touchscreen monitor and a webcam, the uBot-5 can be used for social telepresence applications. Additionally the teleoperator can use the uBot-5's manipulation capabilities to perform useful work.

The uBot robot is specifically designed to operate in human-centric environments and it has been shown capable of lifting, pushing, and throwing tasks. The video below shows the robot picking up a ball from the ground while remaining stable.



For a nice overview of the history of the uBot series of robot, you should read this paper (pdf).

ZMP e-nuvo WALK is now MSRS powered

A good number of academics are still opposing Microsoft's Robotics Studio software for robot development, but robot manufacturers have welcomed the software giant's entry to the robotics market. Many have not lost a second to write drivers for their robots that support Microsoft's technology which has so far proven to be a very solid implementation of a robot control architecture. Recently, the Japanese ZMP announced that in January 2008, they will start selling their new e-nuvo WALK humanoid robot with support for MSRS making it the first of its kind to support Microsoft's architecture. The e-nuvo WALK is only half a humanoid lacking an upper body as you can tell from the photo below. The 5.5 pound robot has 6 motors in each legs and it is specifically designed for use as an educational tool in the classroom; the robot stands 14 inches tall. It will cost a rather steep $5345 when it becomes available for sale online this January (but only in Japan.) Hopefully, the robot will eventually become available in North America.