Artificial Intelligence and Robotics

International Humanoid Robot Olympic Games

2009-11-08T00:51:00.000-08:00

The Chinese have just announced that they intend to hold the first ever International Humanoid Robot Olympic Games in the city of Harbin possibly in June 2010. There are few details about the event other than what was posted in a brief article on CCTV's website; you can read the article here.

The organizers are busy trying to find sponsors for the robot Olympics while they expect 100 Universities from 20 countries will send robots to compete. The Olympics will be exclusive to humanoid robots. The events will be a mixture of traditional Olympic sports such as track and field and more robotics related activities such as doing housework.

The organizers claim that the robot Olympics are designed to help make more intelligent robots for use at home. The Humanoid Olympic Games will be competing with the well established RobotCup competition which attracts much attention every year. Robot rescue and robot soccer are the two main events with the latter having the longest history in robotics competitions (see our previous stories "Robot soccer: Then and now!" and "RoboCup: When and why did robots start playing soccer? for more information".)

I am curious to see if the Humanoid Olympics actually take place and if so how many Universities will be able to take part. More importantly, it is amazing to me that in 2009, we can start thinking about robotics competitions that involve only humanoid robots and expect many participants. It goes to show how quickly the field of robotics has advanced in the last decade.

IBM 7094 sings Daisy Bell

2009-11-03T01:28:00.000-08:00

Artificial Intelligence has come a long way in the last 50 years. But there is still something creepy and yet intoxicating about the IBM 7094 singing Harry Dacre's 1892 classic Daisy Bell.

In 1961, the IBM 7094 became the first computer to sing, singing the song Daisy Bell. Vocals were programmed by John Kelly and Carol Lockbaum and the accompaniment was programmed by Max Mathews.

Boston Dynamics PETMAN humanoid robot

2009-10-27T13:59:00.000-07:00

Boston Dynamics makers of BigDog, i.e., the coolest robot ever made, are in the process of building a dynamically balanced humanoid. PETMAN is their new 2-legged robot sponsored by the US Army for the purpose of testing chemical protection clothing for the soldiers.

Boston Dynamics has not released much information about PETMAN other than that the project has 2 phases : phase 1 has a 13-month duration focused on design and phase 2 has a 17-month duration for building, validating and installing the robot. The company expects to deliver the robot in 2011.

If you care to know how impressive PETMAN truly is, check out the video below released by the company. It shows the humanoid robot walking on a treadmill using principles similar to how a human walks; it also maintains balance when large external forces are applied to it. If it was me, I would not anger this one by pushing it. It makes Toyota's running humanoid look like a toy. For what it's worth, however, it doesn't look like PETMAN can actually turn but maybe this is one upgrade the US Army hasn't paid for yet.

Some more information on PETMAN by Boston Dynamics can be found here.

Microsoft SenseCam to be mass produced

2009-10-25T00:01:00.000-07:00

Scientists at Microsoft Research Cambridge have for years been working on SenseCam, a wearable device that records at regular intervals photos of its user's environment. The camera comes with enough memory to store up to 30,000 VGA resolution images along with the necessary tools for the user to browse the data using her computer. The VGA camera is not the only sensor on the device. According to the project website,

SenseCam also contains a number of different electronic sensors. These include light-intensity and light-color sensors, a passive infrared (body heat) detector, a temperature sensor, and a multiple-axis accelerometer. These sensors are monitored by the camera’s microprocessor, and certain changes in sensor readings can be used to automatically trigger a photograph to be taken.

The main goal behind SenseCam is to help those with memory problems to remember important events. People suffering from Alzheimer's disease have been shown to benefit from the use of SenseCam.

The following video gives a brief explanation of how SenseCam works and how it is meant to help those with memory problems (keep in mind that the video may be a couple of years old but the information is still correct.)

A UK company, Vicon, has recently licensed the technology and they have committed to mass producing the device. According to Microsoft Research there is a huge demand for SenseCam which they could not satisfy. According to the Press Release, Vicon unveiled their SenseCam product at the annual Neuroscience Conference in Chicago on October 17-21. I believe the new product is called the ViconRevue.

I really hope that this product is only used by those who truly need it and it is not turned into another toy for the Twitter generation; the last thing I want is to be surrounded by people who have no sense of privacy taking my photo everywhere I go.

MIT's Micro Aerial Vehicle (MAV)

2009-10-19T04:37:00.000-07:00

A team of students from MIT's Computer Science and Artificial Intelligence lead by Prof. Nick Roy developed an autonomous Micro Aerial Vehicle capable of autonomously mapping and navigating a complex environment.

The team built upon a quadrotor helicopter developed by Ascending Technologies which they equipped with a Hokuyo laser rangefinder, small USB camera, and a micro-controller for interfacing with the sensors and actuators while providing communications to off-board computers for additional number crunching.

The team participated in and won the AUVSI International Aerial Robotics Competition (IARC) which was held on the campus of the University of Puerto Rico at Mayagüez (UPRM.)

More information about how the MIT-MAV works in the video below.

On the security and privacy of future household robots

2009-10-17T17:15:00.000-07:00

It is obvious to most people that a robot is to a large degree a digital computer on wheels (or legs more recently.) It has sensors and actuators that allow it to observe and manipulate its environment either autonomously or under remote control. Household robots such as the very popular Roomba have sold millions and it is predicted that this number will increase vastly in the coming years. Many if not all of these household robots also come with built-in functionality that allows them to be controlled remotely via an Internet connection; this implies both receiving motion commands and also transmitting sensor data to the remote user's computer.

We all know that computer security is a major hassle for all computer owners. Literally seconds after a computer is connected to the Internet it comes under attack from hackers trying to gain illegal access and steal personal information or use the machine for other illegal activities. Since household robots today all have on board computers connected to the Internet, it makes sense that hackers might try to obtain control of them. In other words, a hacker could potentially take over your household robot and use it to take photos and video of its unsuspecting owners. If future robots are larger and more dexterous, then a hacker could potentially use it to cause property damage. Household robot security is a serious issue that robot manufacturers have to start thinking about; consumers should also demand that robots are designed to be secure.

Researchers at the University of Washington recently published a paper that presents the results of a study on the security of household robots currently available in the market. They focused on studying the 2008 versions of Robosapien V2, Rovio, and Spykee.

The researchers found a number of vulnerabilities that would allow any savvy hacker to take control of the robots or intercept sensor data transmitted over the Internet. In many cases, the data such as video and audio is transmitted unencrypted which means a man-in-the-middle attack would allow anyone to receive it. In addition, some of the robots allowed remote users to login without encrypting the username and password information allowing hackers to steal both with ease. For what it's worth, Robosapien V2 was the most secure of robots simply because it does not connect to the Internet; however, it is still vulnerable to attacks by people using off-the-shelve infrared remote controls.

The vulnerabilities identified in this study are disturbing especially if one considers that at any moment someone could be spying on you and/or your children without your knowledge using a device that is to a large degree a harmless toy. The researchers suggest that household robot manufacturers should place more emphasis on security issues in order to protect consumer privacy. The message is clear that whether one considers the basic household robots available in the market today or the much anticipated and more capable future robots, security is a design parameter that should not be second thought for the robot designers.

You can retrieve the complete paper from one of the authors website here.

PhotoSketch: Taking image composition to the next level

2009-10-06T13:48:00.000-07:00

We don't often write about work in Computer Vision but when we do we always present something that is likely to blow your mind (see VideoTrace and Adobe's Interactive Video Editing.)

Enter PhotoSketch, work that was recently published in ACM's Transactions on Graphics. The application takes as input a rough, hand drawn sketch of a scene and then generates a realistic image using information found on the Internet. The results are magnificent to say the least considering the difficulty in filtering the vast amount of noise in online images. The application requires that the user supplies labels with each object sketched; these are used to search for relevant images online. The object outline along with automated segmentation algorithms are used to extract specific objects from the photos found, e,g, a dog, an elephant, a car, a Frisbee etc. Another mechanism is used to blend the segmented images together and generate appropriate shadows. The results are jaw dropping to say the least considering how little input from the user is required.

The video below explains the process in more detail.

The Precision Urban Hopper military robot

2009-09-16T13:12:00.000-07:00

Looks like Boston Dynamics makers of the very well known Big Dog robot are at it again. A recently released video shows a new, small, 4-wheeled robot that can leap over obstacles of up to 7.5 meters high. The robot which can be seen in action in the video below, utilizes a piston-actuated leg to jump over obstacles that are up to 60 times its own height. The intended use for the Precision Urban Hopper is obviously military surveillance in urban environments. According to news reports, the robot will be available for use in the field next year. Lastly, the robot is meant to be remote controlled and not fully autonomous. I do suspect, however, that the final model will most likely allow for some autonomy to assist the human operator. This is almost always the case with such robots.

Nao Robot plays with Microsoft's Surface computer

2009-09-10T02:12:00.000-07:00

I have been a bit busier than expected the last couple of weeks and posting on this blog has been slow. So, I figured I make it up to you by posting a video of Aldebaran's Nao humanoid robot manipulating some photographs displayed on Microsoft's Surface computer.

This is another nice promotional video for the French robotics company. Aldebaran Robotics are making a serious effort to market Nao to academics and at the same time get the average person excited about the many possible uses of such dexterous, intelligent, humanoid robots. If you want to see another video of Nao walking and manipulating objects, check out our previous post "Introducing Nao."

The Lumiere project: The origins and science behind Microsoft’s Office Assistant

2009-08-31T14:59:00.000-07:00

I think by now just about everyone has heard about the most annoying feature to have been included in a commercial piece of software. I am talking, of course, about Clippy, the Microsoft Office Assistant that many loved to hate. Clippy was first included in the 1997 release of the Office suite and continued to be part of the product line until 2007 when it was permanently removed.

Many people know Clippy as a major nuisance but few know the story behind the technology and why it sucked so much. Keep reading because I am about to tell you all about it (if you don’t want to read the whole story, you can just jump to the video at the middle of this post; if you watch the last minute you will get a glimpse of Clippy’s grandfather.)

Origins

In 1993, Microsoft researchers from the Decision Theory & Adaptive Systems Group established the Lumiere project to study and improve human computer interaction using Bayesian methods. The group wanted to create smart technologies that can observe a user interacting with a computer program and infer his/her goals and needs providing valuable feedback and assistance as necessary. Developing such a technology makes sense since many people often become intimidated by complex software interfaces. I won’t bore you with the details of what Bayesian methods are and why they are good. The mathematics behind such methods is solid and has had many useful applications to date.

So, is this an easy problem to solve?

Actually, inferring a user’s intent is a very hard problem no matter how good your math is. The Microsoft team had to infer user intent from his interaction with the program, e.g., mouse movement, what menu items were selected, context (what is the user trying to do – remember how Clippy always came up saying something like “I think you are trying to write a letter. Would you like some help?”) and specific text queries by the user, e.g., how do I print a document?

Any user model that can adequately capture all the relevant information will necessarily have many variables. The values of these variables must be estimated over time. Moreover, different users tend to interact with a piece of software differently. For example, an experienced user is most likely to need less help; the same user may also help with the more obscure features of the software compared to a novice. Personalization is a very important factor in ensuring that such systems work well.

To make a long story short, the Microsoft researchers led by the senior scientist Dr. Eric Horvitz were making good progress and in 2 years time they already had a nice system working. So, in 1995 and as the team had already started collaborating with the Microsoft Office production team, they put together a demonstration of Lumiere’s inference engine for Excel. The video below is a 9-minute tour of Lumiere working in Excel. In the video, Horvitz explains how the inference engine worked in 1995 and how they envisioned it working in later versions using a cartoon character front-end. Watch the last minute of the video for a glimpse of Clippy’s grandfather.

After the video, I explain using evidence from a number of Microsoft Research publications and personal knowledge why Clippy worked so poorly in the 1997 release of Microsoft Office.

Clippy debuts

The Bayesian inference engine demonstrated in the above video works like a charm monitoring the user’s behavior, inferring his intent, and providing help in a contextual and personalized fashion.

Two years after this video was recorded and after much collaboration between the research and product teams, the Lumiere project debuted as a well advertised feature of Microsoft Office 97. Clippy was one of the many cartoon characters that were available as the engine’s front end interacting with the user.

Unfortunately and as we all know, Clippy worked so poorly that it was not long before users started complaining about its behavior. So, what went wrong?

The reasons behind Clippy’s massive failure

Well, after doing some research I found out what went wrong. In a paper published in 1998 at the Conference on Uncertainty in Artificial Intelligence (UAI), the Lumiere team described the inner workings of the Assistant’s inference engine and also how much of it was included in the released version of Office 97. Below is a list of the features that were excluded from the product release (those keen enough can cross reference the list with what was demoed in the video above.)

No persistent user profiles.

No reasoning about user competence, i.e., novice versus experienced user
Small event queue with emphasis only on the most recent interactions of the user with the software interface (this means the engine was trying to guess the values of many variables using very little data.)

Separation between user interface events and word-based queries; for word-based queries the engine ignored any context and user actions.

Last and possibly most important and I quote from the paper, “The automated facility of providing assistance based on the likelihood that a user may need assistance or on the expected utility of such autonomous action was not employed.” Instead, “The Office team has employed a relatively simple rule-based system on top of the Bayesian query analysis system to bring the agent to the foreground with a variety of tips.” This is why Clippy kept popping up all the time. It was not using the mathematically correct engine that the researchers had designed. It was driven by some rule-based system that one or more of the developers from the product team thought was a reasonable substitute.

Why did Microsoft cripple Clippy?

Obviously for some reason many of the features in Lumiere’s Bayesian inference engine never made it into Office 97.

Why?

I have not been able to find an official document that explains why most if not all of the inference engine’s features were not included in the Office 97 release. However, I can provide some informal evidence based on personal knowledge.

Some time in 2000 or 2001 when I was still a graduate student, Dr. Horvitz gave an invited talk at my university. He talked much about his HCI research and the Bayesian modeling techniques he had been studying for years. A question about Clippy was eventually and unavoidably asked. What the heck happened with that?

I recall his response being that what happened was as noted earlier much of their careful mathematical modeling of users never made it in the final product. He explained that the reason for this was a lack of disk space. You see, the Office suite ended up being much more bloated than originally expected and since most of the more mundane features were considered essential the product team decided to limit the amount of space available for the Office Assistant component. This is why so many features had to be removed. They did not have space for it all.

This is the story of the Microsoft Office Assistant or Clippy as it is most widely known. Microsoft discontinued the Office Assistant (more accurately turned off the feature and I very much doubt that anyone bothered to turn it back on) with the release of Office XP on 2001 and so Clippy is now resting in peace somewhere in a backup drive in Redmond.

The courageous story of Clippy and the tragic story behind its humble beginnings, rise to fame, and downfall are now part of history. But at least, you now know the truth of the reasons behind its unfortunate demise.

LEGO Mindstorms robot solves Sudoku puzzles

2009-08-27T03:55:00.000-07:00

How about a nice little robot build using the LEGO Mindstorms kit that can autonomously solve a Sudoku puzzle? Hans Andersson who lives in Sweden constructed such a robot in his spare time using a robotics kit that he initially purchased for his two kids.

The result is fascinating. Hans used a light sensor to read the puzzle using some clever image processing techniques. He wrote an optimized algorithm for solving the Sudoku puzzle and he, lastly, equipped the robot with a pencil to fill in the numbers.

Very nice work. Enjoy the video below that shows the robot perform its task.

PS: Keep an eye on Hans' website because he will eventually post instructions on how to build your own Sudoku solver robot.

KUKA Light -Weight Robot (LWR) goes on sale

2009-08-26T04:14:00.000-07:00

Industrial robot powerhouse KUKA Robotics engineers have finally finished developing their new Light -Weight Robot (LWR) which is now officially on sale. Don’t ask me for the price because I don’t know. The LWR is a very capable robot that will find many uses in industrial applications and research laboratories.

According to the information that KUKA released about the LWR, some of its features include precise position, torque, and impedance control; torque sensors in every one of 7 axes; smart and safety conscious design; low weight (about 15 kg); and low power consumption. The robot has a nominal 7 kg payload. The KUKA LWR also comes with all the necessary software so that it can be programmed with ease. Moreover, an engineer can teach it to perform a repetitive task simply by demonstration, i.e., by moving the robot manually through each step of the task in the correct order.

The video below shows a demonstration of the KUKA LWR climbing up and down a ladder with ease.

AIspace: Learn artificial intelligence using this collection of freely available interactive tools

2009-08-23T03:33:00.000-07:00

A few days ago, I wrote about what I believe is the average person's irrational fear of robots. As some people pointed out, some of these fears exist because we don't really understand what Artificial Intelligence is and what it is capable of. So, I thought that some/many/all of our readers may benefit from learning the basics of AI.

To that effect, I would like to bring your attention to AIspace which is an online collection of interactive tools specifically designed for teaching introductory AI concepts to students. The tools have been under continuous development since 1999 at the University of British Columbia, Canada.

AIspace includes interactive tools for learning about arc consistency for constraint satisfaction problems, belief networks, graph searching, decision networks, neural networks, planning, and robot control. The tools are also accompanied with plenty of documentation to help anyone get started in AI.

The video below provides a short overview of AIspace. For the record, the person narrating in the video is Prof. Alan Mackworth who among many other honors is also the Founding Father of Robot soccer, i.e., Robocup.

Our fear of robots

2009-08-16T01:32:00.000-07:00

Last week, I wrote about Toyota's humanoid robot and its new running skill. The post received a large number of visitors as it was posted on Slashdot and referenced by a number of other technology blogs. The video on YouTube received more than 600,000 views and over 800 comments. That's great and thank you all for watching. However, I was astonished by the comments posted both on this blog and on YouTube. Most of the comments were expressing fear about the robots taking over the world and killing all humans in the process. References to popular movies such as Terminator and Transformers were plenty.

Why are we afraid of robotic technology so much?

Most people have little understanding of what it takes to develop working robots whether humanoid, 4-legged, tracked or wheeled. It is incredible that in the span of just a few decades we have been able to construct machines that walk on two legs, manipulate objects using dexterous hands, can perceive their surrounding environment, and interact with people. Compare that to nature that took millions of years to evolve humans.

That said, these marvels of technology are designed to perform very specific tasks. The robots are not self-aware even though we might think they are because of their human traits. Robots work tirelessly in factory assembly lines performing jobs that require precision and speed. The Japanese are investing much money in the development of partner robots because they are predicting a huge shortfall in qualified workers due to a quickly aging population; who will take care of the elderly if there are not enough men and women to do the job? Some robots are designed for use by the military including delivering ordnance but these are always under supervision and remote control. Military robots do not make decisions on who to attack and when but instead follow a well laid out plan given to them by humans.

Will one day robots become self-aware and decide to eliminate humans?

If we are to believe movies, TV, and science fiction novels, then chances are that yes sooner than later advanced, intelligent robots will enslave humanity. I am not so sure why but apparently they will. To me it seems the doomsayers take the stance that robots will be very much like humans leading to a disastrous outcome (for us, of course.) In the course of history humans have not given a second thought to enslaving other humans whom they thought inferior to them. If robots are smarter and more capable than us then surely they will inevitably decide to enslave us.

Not necessarily say I.

There is no reason why self-aware robots will develop to be evil and hate humanity. There is a good chance that the robots may decide to help humans. Along with becoming self-aware, they may also become sympathetic to us. Robots do not feel tired and can easily be repaired if necessary. Not to mention that if the robots develop an adventurous spirit inherent to humans, then chances are that they will move away from the Earth exploring the vastness of space something that humans will not be able to do for a very long time.

Anyways, to conclude a longer than usual post, I don't think that robots will necessarily turn bad and try to enslave humanity regardless of what you watch in the movies. We are designing our robots to co-exist with us and we are making efforts to make sure that in the future robots will be treated ethically just like we treat other humans. There is nothing to fear. We should be proud of our achievements in robotics and artificial intelligence. Maybe one day if humanity perishes and robots inherit the Earth, they will write their own version of Genesis with a human being the creator who made them in his image.

Touchable holography

2009-08-08T16:46:00.000-07:00

Touchable holography may just be the next big step for holographic displays. Over the years, there has been considerable progress in developing holographic displays but none of these allow an immersed user to truly interact with the displayed objects because touch is missing.

Until now, that is!

Researchers from The University of Tokyo Shinoda lab have figured out a way to add touch to holograms.

First, they utilize an LCD-based, concave mirror, holographic display for displaying the 3D images. Using this display, images appear to float 30cm away from the display surface. Second, they use a couple of Wii controllers for hand tracking in order to determine the user's hand location with respect to the displayed objects. Third, they provide tactile feedback using their own Airborne Ultrasound Tactile Display (AUTD.) The device uses acoustic radiation pressure to induce tactile feedback as users interact with the displayed objects.

The touchable holography prototype will be on display at the 2009 SIGGRAPH conference in the emerging technologies track.

The video below explains how touchable holography works.

Microsoft's pressure-sensitive keyboard prototype

2009-08-08T16:24:00.000-07:00

Microsoft's hardware division has unveiled a new prototype keyboard which is pressure sensitive, i.e., it doesn't only record which key you press by also how hard (8-bit resolution).

I never thought of a pressure sensitive keyboard before but watching the video below demonstrating its capabilities and possible uses definitely makes me think that it is a peripheral device I might want to own some day. I really like how it can be used to determine incorrect key presses, a functionality that should be of great use in mobile devices with tiny keyboards. Students participating in the first annual Student Innovation Contest in Victoria, Canada, will be the first to play with this keyboard trying to come up with innovative interfaces that take advantage of its pressure sensitivity.

Here is the video from Microsoft Hardware explaining how the pressure-sensitive keyboard is constructed and how it can be used to interact with a computer.

Contest winners will be announced on 6 October 2009. I can't wait to see what kind of killer applications the students come up with.

HAL exoskeleton taken for a walk outside

2009-08-06T02:28:00.000-07:00

I've written twice before on the Cyberdyne HAL suit, one of the first working exoskeletons to be demonstrated and brought to market. You can read the below articles for a description of the origins and capabilities of this exoskeleton.

HAL exoskeleton available for renting in Japan

Japanese quadriplegic attempts mountain climb using robotic exoskeleton

Recently, for what I am guessing is a marketing effort, 3 Cyberdyne employees put on their HAL suits and went outside for a stroll. You can watch that in the video below, courtesy NECN.

Numerous advancements in exoskeleton technology were achieved in 2008 with many research labs and companies releasing information about exoskeletons that are only one or two years away from going to production. This is an exciting area to keep an eye on because of how useful such exoskeletons can be for the elderly and the disabled.

Nearest Tube iPhone augmented reality application

2009-08-04T02:39:00.000-07:00

Acrossair has developed a new augmented reality for the iPhone 3GS designed to help iPhone owners find their nearest tube station in London. As you can see from the video at the end of this post, the application (which is still under review by Apple and not available on the App Store) utilizes the phone's compass and GPS localization to display the directions to the tube stations near the user. This information is displayed over the video captured with the phone's build-in camera to make consuming the information easier for the user.

Clearly, the application does not perform any image processing such as structure from motion computations. It also does not embed 3D objects into the scene as most would expect from an augmented reality application. I don't know what the exact definition of augmented reality is since it is not area of research of my own but I usually expect more than just an overlay of data over a video stream. That said, the iPhone and all other smart phones in the market are not powerful enough to perform such computationally expensive computations so the current crop of augmented reality applications are likely to be more like Acrossair's Nearest Tube.

Finally, let me remind you of an older article of ours discussing Nokia's efforts to bring this kind of functionality to smart phones more than a year ago. Interested readers can read our article With computer vision by your side you will never have to ask for directions again. The same post includes a research video from the University of British Columbia showing how with image processing, i.e., structure from motion estimation, objects can be added seamlessly into the video stream.

And so, here is the Nearest tube demonstration video.

Augmented reality business card

2009-07-30T02:45:00.000-07:00

Check out in the video below, a nifty augmented reality business card created by James Alliban. In the below video, James shows a business card with a distinctive pattern on it to a computer equipped with a cheap web camera and the software automatically renders a simple 3D object (attached to the card as it moves around in 3D) and plays a short video about the person whose business card is shown. A nice idea if you want to convey more information about yourself than can possible fit on a paper business card.

For another interesting augmented reality application relating to smart phones and direction finding in an urban environment see our older post titled "With computer vision by your side you will never have to ask for directions again".

New leader for the Netflix Prize

2009-07-26T01:35:00.000-07:00

In a last second twist, a new team has managed to best the previous Netflix Prize winner BellKor's Pragmatic Chaos ready to claim the million dollar top spot. The new team called The Ensemble is the amalgamation of 3 other teams known as Grand Prize Team, Opera Solutions, and Vandelay United.

To remind you of the situation which you can also read in my previous post here, BellKor's Pragmatic Chaos was the first team to submit an algorithm achieving more than 10% improvement on recommendation accuracy. The exact improvement was 10.08% on the benchmark data set provided by Netflix. When this happened some 29 days ago, a timer was started giving other teams 30 days to do better in order to claim the prize for themselves. At the conclusion of the 30 days, the highest scoring team gets the money.

In an incredible twist, just a day before the 30 day period expired, The Ensemble team submitted an algorithm that achieves 10.09% percent improvement. This means, that with one day remaining, BellKor's Pragmatic Chaos has to either improve their results by at least 0.01% or finish in second place behind The Ensemble.

What a great finish!

Let's see what happens in the next 24 hours but in my opinion, Netflix should compensate both teams. Or, in the least, they should give everyone another 30 days to respond to the most recent challenger. Why not?

You can follow the most recent developments via the Netflix Prize leader board here.

Astrobotic Technology’s lunar robot

2009-07-23T01:55:00.000-07:00

After finishing first in the DARPA Urban Challenge and second in the Grand Challenge (due to an unfortunate mechanical failure,) CMU's Dr. William Whittaker is ready to conquer the Moon with his Astrobotic Technology's rover entered in the Google Lunar X Prize.

The Lunar X Prize offers a top prize of 20 million dollars to the first team that lands a small robotic rover on the Moon and beams back video as it navigates on the Moon's surface for at least a distance of 500 meters. The hard deadline for completing the mission is set for December 2014. A total of 19 teams from around the world are currently registered for participation.

Astrobotic's team announced today that they have solved a major problem for the Moon rover. Temperatures at the surface of the Moon during the day reach 270 degrees Farenheit; designing a robot that can survive such extreme temperatures is not an easy task. Astrobotic's Red Rover will use a smart design and navigation method to survive the harsh conditions. More specifically,

Noon at the equator is hotter than boiling water: 270 degrees F. The robot beats the heat by keeping a cool side aimed away from the Sun to radiate heat off to the black sky. It travels toward or away from the sun (generally east or west) without turning its radiator into the light. Only the solar cells on the hot side ever face the sun. The robot can travel north and south by tacking like a sailboat.

This sounds like a neat solution to a hard engineering problem. Red Rover is supposed to use this technique when it lands on the Moon in 2011 with a mission to explore the landing site of Apollo 11. Incidentally, this past week also marked the 40th anniversary of humans walking on the Moon's surface. NASA astronauts Neil Armstrong and Buzz Aldrin were the first of a total 12 people to walk on the Moon. It looks like number 13 will be a robotic rover.

Nice video showing Red Rover from a media photo shoot below.

Toyota’s running humanoid robot

2009-07-20T02:52:00.000-07:00

We all know of Honda’s ASIMO humanoid robot and its amazing walking and running capabilities. Other research labs though are not far behind developing robots just as capable. In fact, Toyota has an excellent and very advanced robot in the making even though they entered the game much later than Honda.

Toyota’s most recent humanoid robot prototype (one of many partner robots the automotive giant is developing) stands 130cm tall and weighs 50Kgr. Its legs have 7 degrees of freedom and it can run at an average speed of 7 km/h. In contrast, ASIMO’s maximum speed is 6km/h. The Toyota researchers had to develop new real-time methods for balance control. These methods make it possible for the robot to remain balanced when an external force such as a push from a human is applied when in motion.

The below video from Toyota demonstrates the running capabilities of the new humanoid robot. The robot takes a step every 340ms and has no contact with the ground for 100ms of that. Notice in the video how the robot remains balanced even after pushed by the human.

Finally, even though this new robot is impressive, it is still limited on moving over flat surfaces and it can only recover from small external forces. But then again, even Rome was not built in one day!

Reinforcement Learning introduction by example

2009-07-15T06:20:00.000-07:00

A group of McGill students have created a brilliant short video introducing the basic ideas behind Reinforcement Learning (RL) and one of the most popular RL algorithms known as Q-learning. Using a hypothetical bartender robot named Shaker, the video explains how an agent learns to act from interactions with his environment and a reward/punishment system.

At first, such agent thinks that all actions are equally good; this is a consequence of the fact that he has no prior experience that would allow him to make a proper action selection. As a result, the agent chooses actions at random for all situations (also known as states.) At the conclusion of each action, the agent receives either a reward or a punishment from the environment. The former denotes an action that was a good choice for the particular situation while a punishment denotes the opposite. Continuing in this fashion, over time, the agent learns which strategies, i.e., sequences of actions, help him maximize rewards or minimize punishment. The learned value function allows the agent to act rationally in all situations.

The simplest Reinforcement Learning algorithm is known as Q-Learning. It is a model-free method since a model of the world is not available to the agent a priori. The Q function that the agent learns interacting with his environment gives a value for each situation and action combination possible. Even though Q-Learning is a simple and yet powerful algorithm, it is not a practical one. The number of states and action combination that the Q-function must be learned for is often large if not infinite. An agent will often fail to explore all cases unless considerable amount of time is made available to him; by the time your average robot learns to act using this method, you and everyone else on Earth will most likely be living on Mars. There are many more powerful algorithms that researchers have invented over the years that tackle some of the above issues; I will discuss some of them in future posts. For now, remember that we still have a long way to go before we have robots that efficiently learn from experience but progress is continuing at a fast pace.

If you found the above textual introduction to Reinforcement Learning boring or difficult to follow, then the below video which is also the main point of this post might clear things up. Enjoy!

RoboCup Rescue 2009 photos

2009-07-14T01:44:00.000-07:00

Amir, one of the participants in this year's RoboCup Rescue competition was kind enough to let me know of a large collection of photos from the competition that he would like to share with all of us. For context, the rescue competition is designed to stimulate research in building autonomous or semi-autonomous mobile robots that can assist in rescue operations in disaster zones, e.g., finding survivors in a collapsed building after a major earthquake. The competition has been taking place annually for more than a decade and I can tell you that the rescue robots are becoming more advanced every year. I am very optimistic that in another 10 years such robots will become a rescuer's best friend helping save many lives.

With Amir's permission I have included a couple of the photos at the bottom of this post, but if you want to see the entire collection of more than 100 high resolution photos, go here.

Thank you Amir and I hope your team was successful at the competition.

Wiimote control of a 15-tonne machine

2009-07-12T01:28:00.000-07:00

If there is anything Australians know how to do well, it is mining. Transmin engineers had some fun building a steel monster controlled using a Wiimote. It is not something that you will be able to buy for home use (not that you would ever want one of these machines for home use) but it is a fun little project attached to a very serious project of building a massive piece of equipment for mining. Watch the video and share in the fun.

Thanks Adrian for the heads up.