Freeze frame demo by Point Grey Research

Vancouver-based Point Grey Research famous for their digital imaging products including stereo and spherical vision cameras were busy at SIGGRPAH 2007 demonstrating their new freeze frame application. Freeze frame creates a visual effect similar to the one popularized in the Matrix movies; this can be seen in the animated GIF at the top of this post. The application is a demo for the company's hardware and software for synchronizing a large array of Firewire-based cameras streaming 320MB/sec of data to a single PC.

32 2.0MP IEEE-1394b Grasshopper cameras (model GRAS-20S4C) cameras are connected to the system via Point Grey 3-port 1394b hubs and daisy chaining in groups of 4 cameras. The four hubs provide additional ports and power to the system, while daisy chaining maximizes the reliability of the system and minimizes cabling requirements.


On the host side, a single PC contains two Point Grey dual bus IEEE-1394b PCI Express cards. Each PCIe card uses two (2) LSI Logic 1394b physical layer chips to allow simultaneous image acquisition from up to 16 cameras. With each 1394b bus effectively supporting 80MB/s, and using the 2.5Gb/s single lane (x1) PCIe serial link, the total data throughput of a system with two dual bus cards is maximized at 320MB/s!

You can also watch the demo in action in the following SIGRAPH video by Viral; the freeze frame demo is shown towards the end of the video.


Online Videos by Veoh.com

Freeze frame demo images are copyright Point Grey Research.

Edible robotics

Can you construct a robot's body using produce, cheese, rubber bands, and paperclips and then actuate it using an electric motor? And after you have impressed everybody with your creativity finish by eating your robotic creation? Apparently, you can. Check out this fun project called edible robotics including a cucumber-based robot hand and a Rice Krispies-based humanoid robot. The group's mission statement explains,

Edible Robotics represents a sea change in thinking and action regarding our relationship with our friendly robot servants. Our goal in exploring this new territory is to report back from the frontier with our discoveries and observations in the hopes that others who follow will not fall prey to pitfalls we've stumbled across in the interest of science.

Video evidence follows,

Robotics engineers get paid well says CNN

If you are considering a career change or if you are now entering college and have to choose a major then consider becoming a robotics engineer. CNN performed a bit of research to try and find the 5 areas where demand for talent is higher than supply and they found that robotics engineers are in high demand. You can easily receive a 6-figure salary today if you have a few years of industry experience. And considering that everyone is predicting that the robotics (especially consumer robotics) market will soon explode, getting a degree in robotics and artificial intelligence is probably a good idea.

Holographic display at SIGGRAPH

Recent research presented during the recently concluded SIGRAPH conference focused on the design and implementation of affordable 3D or holographic displays as are more commonly known. A team from the University of the Southern California, Fakespace Labs and SONY presented a paper on their autostereoscopic, omnidirectional and multiview 3D display. Autostereoscopic basically means that you do not need any type of special apparatus such as glasses in order to see the rendered images in 3D. Omnidirectional means that the models rendered look good from all directions. Finally, multiview entails that multiple people can be watching the display and seeing a correct rendering of the image from their point of view.

The title of the paper is “Rendering for an Interactive 360º Light Field Display“; the abstract explains the focus and contributions of the work,

We describe a set of rendering techniques for an autostereoscopic light field display able to present interactive 3D graphics to multiple simultaneous viewers 360 degrees around the display. The display consists of a high-speed video projector, a spinning mirror covered by a holographic diffuser, and FPGA circuitry to decode specially rendered DVI video signals. The display uses a standard programmable graphics card to render over 5,000 images per second of interactive 3D graphics, projecting 360-degree views with 1.25 degree separation up to 20 updates per second.

I am not an expert in 3-dimensional display technology but this work is exceptional. You can download the paper and/or view a movie of the display here.

Face recognition for computer security

If you are looking for a fun application of face recognition, you should check out Banana Security's new desktop security system codenamed Banana Screen. The software, which is free to download, requires a web camera. Once installed, using a simple procedure, the software constructs a model of the user's face. When the person moves away from the computer, Banana Screen automatically locks the machine. Face recognition is used to unlock the machine without the need to enter a password via the keyboard. The software prevents others from using the computer because their faces are not likely to match the original user's stored face model.

The software is undergoing beta testing so don't expect it to work perfectly. At the moment only Windows XP is supported but the developers are promising that a Mac OS X version is forthcoming.

Make sure that you have a good web camera before attempting to use this software. I tried it using my almost 8-year old Logitech camera which has terrible image quality and the face recognition did not work; although, the face detection seems to function most of the time. Still, I would not suggest that anyone should depend on this software for protecting their computer; some people in the support forum are suggesting that the system can easily be tricked when shown a life-size photograph of a person's face. Nevertheless, it is a fun little application to try playing with if you have some time to spare.

Robot in The Simpsons Movie

I watched The Simpsons on the big screen the other day and I was pleasantly surprised that it was actually pretty good considering that the quality of the TV series has been going downhill faster than you can say “The Simpsons jumped the shark when...” Anyways, this is a robotics and artificial intelligence blog so what does this cartoon movie have to do with it?

Well, as usual, Matt Groening never misses a chance for some good satire on the latest developments in society, culture, science and technology. There is a robot EOD that makes a brief appearance when the Springfield police must disarm a bomb preventing it from killing everyone and destroying the entire city. The robot having developed human characteristics chokes under the pressure resulting in a hilarious outcome. Let's hope iRobot is not working on a personality module for their PackBot robots. You can enjoy this particular clip from the movie below,

iRobot Roomba 500 series robot vacuum

If you haven't heard it yet (it is reported all over the Web,) iRobot has introduced a new Roomba model. The company has updated the extremely popular robot vacuum by introducing the 5th generation Roomba 500 series of floor vacuuming robots.

There are lots of upgrades included in the new 500 series models. A quick survey of iRobot's website reveals that they are currently selling 3 different models, the Roomba 530, 560 and 570. The upgrades include a new look, better cleaning system, faster recharging, better navigation, voice demonstration and new sensors. It is hard to tell by just looking at it but the Roomba 500 packs lots of what are probably very important upgrades.

Obviously the improved cleaning system is a no-brainer update. At the end of the day, people buy this robot because they want to keep their floors clean not so that they can tell their neighbors about their new toy (although I'm sure many do.) So, making a better vacuum cleaner should naturally be the number one goal for iRobot. I hope that they are also looking into making Roomba quieter; I have one of the first generation robots and it is so loud that I almost stopped using it as it was driving me nuts.

In terms of navigation, iRobot has improved Roomba's ability to vacuum more than one room in an efficient way. Using the new Virtual Wall Lighthouses, the robot can focus on cleaning one room before moving to the next one. The lighthouses work as active infrared doors for Roomba. Whereas in the past, virtual walls would always be on preventing the robot from entering/exiting a specific area, the new active virtual walls turn on/off guiding Roomba from room to room in a more intelligent way. It sounds like a good idea to me.

Moreover, the robot also has new sensors for obstacle detection. Roomba's new forward facing infrared sensors help improve its performance when about to bump into objects. Older models would hit walls and furniture at full speed only detecting said obstacles with the bump sensor. The 500 series utilizes 4 infrared sensors to detect obstacles and slow down the robot early, before a collision occurs. It is an important feature for those worried about the robot damaging their favorite piece of furniture, hurting their pets, or even the robot destroying itself over time.

Other than the improved cleaning and navigation, iRobot has also spent much effort in making Roomba easier to use and maintain. The new voice demonstration system uses speech to help a user through the robot's initial setup; in addition, it provides voice feedback during regular operation as necessary. Another good idea but I hope that the robot will use its voice capabilities with lots of restrain. I really don't want my vacuum talking to me about every little thing happening in its world. The robot's new modular design is also meant to make maintenance easier than before. Finally, iRobot has simplified the onboard interface to improve the robot's ease of use. The higher-end models include an on-board scheduler and a wireless command center.

As I said earlier, there are three new Roomba models, 530, 560 and 570. The higher the model number the more expensive the robot but also the more goodies one gets with it. The 530 is the most basic model that does not come with the Virtual Wall Lighthouses or an onboard scheduler and wireless command center. You have to buy the 560 if you want the first two or the 570 if you want all three accessories. The cost for a 530 is $299.99, while the 560 and 570 sell for $349.99 and $399.99 respectively.

The Tilera TILE64 multicore processor

It would appear that a small MIT startup has beat Intel, AMD and Sun Microsystems to market with a new massive 64-core processor. Tilera Corporation announced a couple of days ago their new TILE64 embedded processor based on their new grid architecture called iMesh.

Tilera Corporation today launched the TILE64™ processor, the first in a family of Tile Processor™ chips based on a revolutionary architecture that can scale to hundreds and even thousands of cores. The TILE64 processor contains 64 full-featured, programmable cores - each capable of running Linux - and delivers 10X the performance and 30X the performance-per-watt of the Intel dual-core Xeon, and 40X the performance of the leading Texas Instruments DSP*. Initial target markets for the TILE64 processor include the embedded networking and digital multimedia markets.

Wow! Tilera is certainly not holding back when it comes to making bold claims about the scalability and performance of their embedded processor design. They certainly brought to market an massive multicore processor well before Intel who announced their own 80-core silicon a few months ago but it has yet to emerge from the lab. It is also interesting to note that TILE64 runs Linux.

Each of the 64 cores on the TILE64 processor is capable of running its own operating system, such as Linux, and provides dramatically more compute performance than any other competing multicore device. Each core is a full-featured, general-purpose processor that includes L1 and L2 caches, as well as an innovative distributed L3 cache. The cores are overlaid with the iMesh network, which provides extremely low-latency, high bandwidth communications between the cores, memory and the I/O.

Apparently any application written for a multicore architecture can easily be ported to the TILE64. I guess this should make programmers' lives much easier. There is also a set of development tools that Tilera makes available to software developers making it possible to more carefully optimize software for the multicore platform.

The TILE64 processor is supported by the most robust set of software tools ever designed for multicore processor technology. The toolset is based on an industry-proven, open programming environment and adds advanced multicore debugging and profiling techniques to accelerate developer productivity and time to market.

It is clear that for the time being this is not a Desktop chip so you won't be able to buy it at the local computer store. Tilera plans to release more versions of the processor with 36 and 120 cores. Let me remind you that Intel's promised 80-core CPU is not due to market for a few more years. The chip is available for shipping at less than $450 for orders of 10K-unit quantities.

Related: Faster computer chips spell good news for artificial intelligence research.

Probo: an intelligent huggable robot

Europeans are working on a new toy robot to be used for therapeutic purposes for children at hospitals. The ANTY project is currently developing their first intelligent huggable robot Probo.

The ANTY project is an interdisciplinary research project aiming to develop a robotic friend for hospitalized children called Probo. The ANTY project will be used to accomplish three important goals concerning different areas of interest. The first and main goal is to develop and build the robot to enlighten the children’s stay in the hospital. A second goal is to use the robot as a multidisciplinary research platform for technical, medical, social and psychological studies. The third goal is to use the project to motivate students and to stimulate new technological innovation.

Probo is a huggable robot based on an imaginary character drawing characteristics from the Proboscidea (Elephant) family of animals. As you can tell from the photo on the left (3D rendering actually,) Probo is cute and very huggable. Researchers claim that Robo will be more than just a toy since the robot is loaded with vision and audio sensors to go along with its artificial intelligence system.

The project is backed by Bill Gates who has already claimed that robotics will be very important in the not so distant future. The Probo team expects to deliver the first prototype in 2008. At first, the robot will be used mainly for research. However, the development team has bigger plans for Probo's future,

Probo will become a fully autonomous intelligent huggy robot. The first prototype is to be expected in 2008, the movements will be limited: the head with trunk, mouth, eyes, eyelids, eyebrows and ears will be actuated. This prototype will be used with hospitalized children and as a research platform to obtain our final goal, to create a friend for life!

Of old computers

This post is guest blogged by Shivashankar B. Nair in response to my earlier request for people to write about their first computer.

The year is 1984, the place a mediocre city (then) in central India - my first encounter with these machines. I had earlier only seen them in magazines. But through the hush hush of the door of a room and through glass panels of a tiny air conditioned room in the Government Engineering College, I first caught a glimpse of what I thought was a computer. Much later I discovered that it was only a microprocessor development system (the HP make [was it the HP68000?]) based on the Intel 8085. I'll talk of it more as I worked on this machine much later.

My first hand-on experience was on the so-called mini computers (HCL machines; HCL stands for Hindustan Computers Limited, an Indian company). It was housed in a big room in a big building! Of course we were not allowed to type our programs on this machine. For this we had to go to another room - the data entry room - where a small microprocessor based machine allowed us to punch in the BASIC program (already written in a coding sheet) sector by sector, track by track, onto those big floppies (no not 5 1/4" but the next bigger & older ones - I forget their size now). . We would then take the floppy insert into those huge vertically mounted floppy drives of the minicomputer and compile the program only to discover numerous errors. So back we go to checking the code & then again queue up at the data entry room for our turn and .... All for just checking whether the 8085 based mini computer could compute the sum of two numbers. Please don't laugh - for that was the very first program we wrote! In six weeks we had written the then so complicated, payroll program in BASIC.

All this was part of my summer internship at HCL office at Chennai (erstwhile Madras). The euphoria of computers was curtailed when I got back to the College for they had none. Computers were real expensive and many did not know how to handle and use them.

When I finished my Masters, I joined a college as a Lecturer only to be bedazzled at the sight a small home computer that when connected to a TV as a monitor could be programmed in BASIC. I remember how any noise in the scan lines on the TV would make the alphabet displayed go zig-zag!

Later I joined the University back at my hometown as a Lecturer and by virtue of that gained entry through that hush hush door to discover the HP68000 MDS. I worked on it for long hours, writing assembly programs. It had an 8085 emulator, an EPROM burner, a printer and a separate Winchester hard disk that could be moved around (the cabinet that housed it had castor wheels). I was thrilled to know that one could see the contents of the registers in the processor in real time. The University then bought an HCL multi-user CP/M based system, that was in 1985. I still remember the funny sounding PIP commands. We could run FORTRAN or COBOL programs. Many a time we would finish a cup of tea by the time the system compiled and linked the code! The editor was a painful line editor. A year later PCs were bought much like the one you cited - two 5 1/4" drives with RAM and no HDD. Ah then we had so many floppies. I still have many of them somewhere at home.

In the later years my Department purchased a PC/XT and then a PC/AT ( 80286 based machine which boasted of a 20MB hard disk). How cautiously I guarded them. The vendor was kind enough to give us a copy of Lattice C and that was my first brush with C. With the book by Ritchie and Kerningham in front, I ventured in to writing my first C program - Hello world! - only to discover numerous errors that scrolled past quickly through the screen. It was a frustrating experience which made me give up on C till much later.

It was in 1986, after attending a workshop on AI, that I migrated to Prolog (then MicroProlog - wonder if anyone reading this has used it). The coming years found me stuck to Prolog for almost everything; of course there were occasions when I used FORTRAN and Pascal for a change. Turbo Prolog was real good. The toolbox that came as a separate entity was excellent. Along with came Turbo C. Yes, I wish Borland continued doing what they did. Today Turbo Prolog has been taken over by PDC Prolog (who now market its enhanced version in the name of Visual Prolog). MicroProlog, I think has now become LPA Prolog.

Somewhere in the early '90s we bought a couple of 80386 systems. Wow, they were real fast (then). We used it for data acquisition, robot control and well what not. To the best of my knowledge those machines are still in good shape. I could afford to buy a "personal" PC only in 1996 costing around 40000 Indian Rupees (roughly $1000). That was huge money then - it still is. (But imagine I could buy a good and much faster laptop now for the same cost!) I used the same for my Doctoral degree - programming real robots in Turbo C and Turbo Prolog and typing my thesis using Wordstar and PageMaker. My nephew and nieces (just 4,4 and 10 years old resp. then) would also use it to play (rough) games - the VGA monochrome monitor was all I could afford! I later gave it to my sister who used it till as late as 2000. It was given off to a company which I was recently informed still uses it for data entry! The systems of yesteryears though slow seem more rugged. Of course MS-DOS was good, transparent, robust and reliable. I quit the University to join a better Institution as an Assistant Professor in 1998 which is when I got to use the Internet using Lynx (text browser).

In recent years there seems to be a rush to update to newer PCs. Almost every 2-3 years (maybe even less) we upgrade the machine to the latest. A couple of times it has so happened that when I just about settled on a machine, they came to tell me to swap it for a new and faster one. Technology seems to be changing so fast that it hardly gives you time to settle on an environment. Things are "improving" though; I have not received an upgrade in the last couple of years but the fear of the "update trend" stopping (for want of funding) is now lurking around. Are the good days over?

Hope you enjoy reading all much more than I did reminiscing them.

Shivashankar B. Nair: I am an Associate Professor in Computer Science and Engineering at the Indian Institute of Technology Guwahati, India, a premier institution funded by the federal Government of India. I have been into teaching and also research for more than two decades. Of course I started of teaching immediately after my Masters degree in Science and continued to do so while I worked towards my Masters in Engineering and later a PhD in AI & Robotics. My current research interests include Artificial Immune Systems, Intelligent Agents and AI & Robotics. I enjoy reading the good ol' Walt Disney and Lee Falks' comics as also watching their videos!

Optimus Prime replica prevents robbery

On the lighter side of news since I have been very busy the last couple of days to research a more scientific story, an Optimus Prime replica robot managed to prevent a home robbery. The Sun reports in an exclusive story (?) that a wooden replica of Transformers leader Optimus Prime constructed by Michael Clarke has apparently scared away a robber.

Robots in Disguise fan Michael Clarke built and carved he towering robot from wood, complete with motion sensors which turn on a stereo when someone approaches.
The imposing model has piecing purple lights for eyes, red and purple neon spotlights.
It stands pride of place in his living room at his house in Gravesend, Kent, but little did he know it would ward off an unexpected guest.

I can't imagine how scared the robber must have been when that huge robot started flashing lights when he/she approached it. Maybe Michael can make a business out of selling his robots as home security systems; after all, we now know that it works!

You can read the complete story at The Sun (with photos of Optimus Prime and his creator.)

Robot pet Pleo marketing

UGOBE is trying to continue building buzz around their new robot pet dinosaur Pleo. They have announced that they will be publishing a series of videos on their website telling us all about Pleo and its makers; they call these webisodes which might just be some new Internet term for an online TV-type series. The email, I received from them a couple of days ago states,

Want to know Pleo better? Now you can watch the first webisode of "Behind the Scenes: The Making of Pleo," a six-part series that tells the story of the creators and the vision that has made Pleo possible.

Check out the first webisode "Pleo and UGOBE's Three Laws" now on PleoWorld (warning: noisy website). Every two weeks from now to October we will release a new webisode for you to enjoy and share. Check back on Friday, August 17th for "Caleb's Tour of UGOBE Labs," the second webisode.

I think it is nice that they want to share with us more details about the making of Pleo. But I wonder, do we really care at this point? The robot went on sale recently but it is not expected to ship until the fall. There is little excitement about the robot to make learning about its design history worthwhile; I can write my biography today but none will care to read it unless at some point in the future I become famous for something. It just doesn't seem like good type of marketing for UGOBE. I find that Wowwee's marketing strategies such as partnering with McDonalds to sell tiny versions of their robots with happy meals much better.

Finally, I have to admit that the more I see videos of Pleo the more I think it is nothing special. It seems very unresponsive. It reminds me of SONY's AIBO that seemed to ignore your every command and just cycle through a few random behaviors; and we all know what happened to AIBO! And I find their insistence on calling this robot a life-form rather annoying. It is not a life-form no matter how many buzzwords you throw at us. A life-form must be able to reproduce and Pleo can't do that (among many other things.)

I would not be surprised if Pleo was a failure but then again, you never know!

Robots are Pentagon's best friend

The US government is really serious about using robots in war zones. Their intentions are really obvious if one considers that thousands of remote controlled robots are already serving in the middle East and that DARPA has put lots of money into the GRAND and URBAN Challenges. News from yesterday as reported by Forbes and Time magazines online are that the Pentagon is looking to purchase another 3000 remote controlled robots for the detection and destruction of Improvised Explosive Devices (IEDs).

As Forbes informs us, nearly half of the US casualties in IRAQ have been caused by IEDs and as such the government is looking at the only option that can truly save lives, i.e., robots. Time magazine says that the Pentagon is going to speed up the process of selecting a vendor because they want 1000 robots delivered before the end of this year. Although several companies supply robots to the military and other law enforcement agencies, I suspect that iRobot with its very successful PackBot series of robots is going to be the company to benefit the most from this; other than the soldiers whose lives will be saved, of course.

JPL's ATHLETE Moon exploration rover

NASA's JPL robotics group has been working hard on a new rover designed for lunar exploration. The All-Terrain Hex-Legged Extra-Terrestrial Explorer (ATHLETE) is one of the coolest robots I have ever seen. As its name reveals, ATHLETE has 6 limbs with dual functions. Wheels at the end of each limb allow the robot to navigate as a rover; but if necessary the limbs actually work as legs turning ATHLETE into a hex-legged robot. JPL's website explains what they have designed the robot for,

ATHLETE is capable of rolling over Apollo-like undulating terrain and "walking" over extremely rough or steep terrain so that robotic or human missions on the surface of the Moon can load, transport, manipulate, and deposit payloads to essentially any desired sites of interest.

The robot is equipped with a large number of stereo cameras that give it a 3D view of its surrounding environment. From what I understood reading the JPL site, ATHLETE would be a remote controlled vehicle although more autonomy is planned for future versions. Interestingly, the limbs also have attachment points for power tools so that they can also be used as arms to perform manipulation and other tasks.

It is hard to do ATHLETE justice using words, so check out the following video of ATHLETE's principal investigator Brian Wilcox describing the robot's capabilities.

ATHLETE photo and video are copyright NASA/JPL.

Urban Challenge teams down to 36

DARPA has completed their site visits with all registered teams for the Urban Challenge and they have selected 36 teams for the semifinals. In late October, the total number of teams will be further reduced to 20 during the National Qualification Event; these last few teams will be allowed to participate in the final Urban Challenge event to be held in Victorville, California at the site of the former George Air Force Base on November 3, 2007. Initially, 89 teams had expressed an interest for participation in the Challenge; last May, DARPA made the first cuts reducing the number of teams down to 53 which were visited and evaluated by DARPA personnel.

Something I noticed looking at the list of semifinalists was that none of the international teams that had registered for the Urban Challenge managed to qualify. This is unfortunate because it was the first time that international teams were allowed to participate in a DARPA event; from what I recall, this was also good publicity for DARPA but I guess if the teams were not up to par and could cause accidents in the last two events, then it makes sense for them to be denied permission to continue. The international teams were from Mexico, Germany, France and Canada. Interestingly, there are 5 semifinalists from California.

DARPA will award $3.5 million dollars to the top 3 teams that will successfully complete the 6-hour Urban Challenge.

DARPA Press Release (pdf).

Update: I was wrong that no international teams qualified for the semifinals. Even though the following 4 teams are listed on the official DARPA document linked above as based on the US, they are actually from Europe. Teams AnnieWay, Berlin, CarOLO and LUX are from Germany. Thanks to Luis Bilik for letting me know of my mistake.

Crows are smart birds

A few months ago, I wrote about an article published in Scientific American discussing work focused on determining how smart ravens are. That study showed empirical evidence supporting the hypothesis that ravens are capable of constructing complex plans using logic. Crows and ravens belong in the same family of birds but the two differ in some basic ways, but both excel in their problem solving abilities. Crows are very smart birds not necessarily outdone by ravens. Recently, I came across a video on YouTube that clearly demonstrates their smarts. The video is about crows living in Japan having adapted to an urban environment. They have developed some very advanced foraging skills; the bird shown in the video below demonstrates a complex method crows use for cracking nuts taking advantage of car traffic. If you watch the video until the end, you will be amazed by how the crows have perfected their behavior to retrieve the food from the cracked nut safely without getting hit by the cars.

Below is another video showing a crow's problem solving abilities. In this case, the bird uses a tool, i.e., length of wire, to retrieve a piece of food that it cannot reach with its beak. Interestingly, the crow knows how to bend the wire in order to create a hook making it possible to grab the food and take it out of its container. Remarkable! I only wish that we had robots that can exhibit the same kind of intelligent behavior.

UltraSPARC T2 open source 8-core processor from Sun

Note: If you don't feel like reading this entire article, then at least read the very end of it. Thank you!

Just yesterday, Sun Microsystems announced the UltraSPARC T2 processor (also known as Niagara 2) under an open source license (GPL) with the chip's complete design to be available at OpenSPARC.net. The new processor has 8 cores each with its own floating point unit and capable of running 8 threads for a total of 64 threads that is double the throughput of its predecessor the T1. And the new microprocessor is open source which I believe means anyone can fabricate it, but honestly, how many can really afford to do that? The fact that the complete design of the chip is available for free will surely make it easier for programmers to port and optimize software for it. I also forgot to mention that the T2 is also power efficient consuming less than 2W per thread.

Sun has termed their new processor as the world's fastest, most energy efficient, true system on a chip and commodity microprocessor in the market today.

Well, that certainly is a mouthful! What Sun's marketing tram is trying to tell us is that if you could actually make this the guts of your Desktop machine then it would allow you to simultaneously surf the Web, finish your homework, rip DVDs, download porn (shame on you,) search for extraterrestrial life with SETI@Home, fold proteins with Folding@Home and get your butt kicked by Polaris in poker. And all that for the affordable price of less than $1000; but only if you buy one thousand units so you better be working on a simulation of the Bing Bang or else you are going to have a hard time getting research funding for the purchase of so many UltraSPARC T2s; or maybe you can enroll in a few more courses with a focus on machine learning and data mining.

In all seriousness, though, the fact that Sun is stepping in a market that has been dominated by IBM and Intel (and I guess AMD to a lesser extent) can only be good news for those interested in scientific computing. It will increase competition and drive innovation along with a drop in prices.

On a similar note, I am really amazed by the incredible amount of computing power we can purchase today for just a few hundred dollars. Even the cheaper dual core systems such as those build around AMD's Athlon X2 microprocessor are vastly superior than anything we had even 10 years ago.

I still remember my first computer which I purchased in 1991. It was an Amstrad PC 1512; the Amstrad PC 1512 was about 5 years old and a bit out of date by then, but that was the only computer I could afford; I worked full-time during the summer of 1991 in order to purchase this computer. It was based on an Intel 8086 microprocessor running at 8MHz and had 512K of RAM. It didn't have a hard drive but it made up for the lack of storage with dual 5 1/4 inch floppy disk drives. Does anyone still remember these? I had so much fun making backups of my programs by copying them from one floppy disk to another and I didn't have to swap disks like those who had machines with only one floppy disk drive. I wrote my first few programs using some interpreted version of BASIC (I believe it was called Locomotive BASIC) that was sold with the machine; other than some early version of DOS, I also had a copy of Gary Kildall's historic CP/M which had a Graphical User Interface well before Microsoft and Apple developed theirs. Finally, the Amstrad almost made me go blind courtesy of its horrible CGA graphics; I was just fortunate that my monitor was gray scale because color CGA with its 16 colors was just plain painful to look at.

And since I am talking about old computers, my second computer was a no-name 386 DX40 PC. It had an Intel processor that could run in one of two speeds at 20MHz or turbo at 40MHz. The turbo button was all the money and I really miss this feature in modern computers. I remember that I had to run the processor at 20MHz for some games because the developers (not the very good ones) had the timing of some games set in such a way that at 40MHz the games run faster and were essentially unplayable. My 386 box had color VGA graphics which was pure awesomeness compared to CGA. I learned to code in C and C++ using this machine and Borland's excellent Turbo C++ IDE; whatever happened to Borland anyways?

After the 386, I moved to a P75 during my undergraduate years and my first 17-inch CRT monitor which I still use after more than 10 years; this is the monitor I use while typing this article. The monitor is a SONY Trinitron Multiscan 200ES and it has served me well. Unfortunately, it looks like it won't last me more than a few more months. Since the P75, I upgraded a few times getting a P90 and eventually a PIII at 500MHz and later a PIII at 1GHZ; the latter was my home machine until about 3 months ago when I purchased an AMD Athlon X2 system for less than $400. And this is the system that I use at home today. We do our research on more recent multi-core CPUs from Intel.

It was nice reminiscing about old computers. I am curious, what was your first computer and what do you use today?

(If you have your own blog, then make a post about your first and current computers linking back to my post/blog and then alert me about it by sending me an email at awesomo2001@gmail.com. I will collect the articles and summarize them in a future post. I think it will be fun for our community to share their own computer (hi)stories in this manner. Just to sweeten the deal, the best post will receive a permanent backlink from this blog. I'm looking forward to hearing from you!)

Space shuttle Endeavour on Photosynth

Microsoft and NASA got together to bring us some incredible views of the Space shuttle Endeavour on the lunch pad just a day before lift-off. The 3D model is part of a new collection for Microsoft's Photosynth technology. Briefly, Photosynth is computer vision software that creates 3D models of scenes using a large collection of images taken from various points of view. In computer vision terms this is called Structure From Motion. Microsoft's implementation is probably the best one that I have seen. Photosynth supports both Internet Explorer and Firefox but only works under Windows XP and Vista; I am not all that surprised that Microsoft isn't going out of its way to support competing operating systems but that's business, I guess. The space shuttle Endeavour collection is complemented by two other collections which include the Vehicle Assembly Building and the space shuttle Atlantis returning to the Kennedy Shuttle Landing Facility after completion of its mission.

I can't wait until Microsoft releases Photosynth to the web allowing us to view more than the carefully constructed in-house 3D models.

If you are looking for a good video introduction to Photosynth's power then watch the following video of Photosynth architect Blaise Aguera y Arcas at TED 2007.

Bandai's visual search engine for music

Japanese Bandai Networks has introduced a new visual search engine that allows mobile phone users to find and purchase music using visual search. The system works in a straightforward manner illustrated in the image below (click it for a larger view.)


Consumers can use the digital camera attached to their mobile phone to take a photo of a CD's cover; this photo is then uploaded to Bandai's servers where using object recognition technology, the CD is identified and information about it is presented to the consumer. If desired, she can then proceed to purchase a copy of the CD via her phone.

Bandai has partnered with music company Label Mobile to digitize 150,000 CD covers. They also use US-based Evolution Robotics' ViPR technology for the visual recognition. ER's toolkit extracts invariant features from the digital images and stores them in a large database. Each object is associated with a number of such features. Given a candidate image to be recognized, the same type of visual features are extracted and matched against those in the database. Each match casts a vote regarding the object that it belongs too. Any object that receives enough votes to stand out compared to the rest is returned as the correct match. The visual pattern recognition system has been shown to work well in many difficult circumstances including under partial occlusion. Evolution Robotics has had much success with this technology as part of the LaneHawk visual scanner for retailers.

I think that it is about time that someone developed a system like this. If it is successful then hopefully Bandai will expand support for additional items such as books. Even better, I hope they bring this service to North America because apparently Bandai is currently only focused on the Japanese market.

Additional information: Bandai Press Release

A Chinese farmer's amazing robots

If you thought that the only way that you could possibly learn how to build robots is to obtain a PhD in engineering then you are wrong. It does help to have some university training but some people are naturally good at applying themselves and building remarkable machines. Take for example Mr. Woo who is a Chinese farmer with no formal training and a large array of walking robots that he constructed over the years. He builds them small and large from scrap metal. The video at the end of this post is a clip from a Paul Merton BBC report. Mr. Woo shows some of his 2 and 4-legged robot creations. Some are small such as the tiny robot dog but others are human size; the latter are incredibly cool as you can tell from the video. The largest of his robots operates as a rickshaw driver. It looks like the ride would be uncomfortable and less efficient than say building a car or motorcycle but hey, it's a robot driving you around!

Mr. Woo lives just 1 hour out of Beijing and apparently is well known so if you happen to be visiting the city then take the time to go see Mr. Woo and his amazing robot creations.

Via Simra.net

The history and ambitious future of Microsoft's robotics team

I was just browsing the online version of the IEEE Spectrum magazine and I came across an interesting article about Microsoft's robotics team. This is the group at Microsoft that has developed (and continuous to develop) the Microsoft Robotics Studio (MRS) development suite. Currently, MRS is produced by a small team of developers with a variety of technical and cultural backgrounds making this an interesting "success" story.

The title of the article is "Robots, Incorporated" which signifies Microsoft's efforts to make a business out of robotics in the not so distant future. As the article explains, MRS was born out of the company's continuous efforts to expand its line of products into new niches that can keep the company profitable in the future.

Today's $11 billion robot sector—mostly industrial robots—will double by 2010, according to estimates by the Japan Robot Association, and it should exceed $66 billion by 2025. Most of the growth will be in nonindustrial applications—especially, analysts say, in areas such as toys, transportation, and health and senior care.

The article outlines how Microsoft entered the world of robotics encouraged by academics who kept asking the company to get involved in educational robotics and upper management (including CEO Bill Gates) who were searching for new applications for some interesting in house projects such as the distributed development framework (among others, of course) that has become a central component of MRS. Interestingly, not too long ago during the International Conference on Robotics and Automation 2007, academics expressed opposition to Microsoft's development platform because of closed standards and source code.

Something I didn't know before but is explained in the article is how Microsoft plans to make money from MRS.

Indeed, future versions may someday find their way into more machines than Windows did—and be just as lucrative. Microsoft's eventual plan is to charge users US $399 to license up to 200 copies of the software components that go into a commercial robot.

In other words, you are very likely to pay a new Microsoft tax for any robot that you buy in the future. Several alternatives to MRS exist including the open source Player/Stage, the commercial Evolution Robotics' ERSP and NASA's newly released CLARAty. Most of these have been around for much longer than Microsoft's offering which is less than 1 year old. And yet, many, if not all, robot manufacturers are already providing drivers for MRS for all their robots.

Maybe Bill Gates was correct last January when he declared that robotics today is at the stage the PC was in the 70s. It is conceivable that the consumer robotics industry will follow a similar path. Some of you may remember how the Windows OS dominated open source alternatives during the 90s when getting an installation of Linux to work on your average PC was a shot in the dark given the limited availability of drivers for many peripheral devices. If every new robot introduced in the next few years comes with MRS drivers from the manufacturer, then what chance do open source alternatives have if support is several months behind Microsoft?

It will be interesting to observe how this battle of robot software will play out in the end. I will be watching with much fascination.

The UK's Grand Challenge for robotics

The British government will be sponsoring a number of teams in their own version of a robot Grand Challenge. In the past, the U.S. Defense Advanced Research Projects Agency (DARPA) has sponsored two Grand Challenges in an effort to create autonomous driving vehicles for supporting troops in military operations. Currently, DARPA is also in the final stages of the first Urban Challenge that is asking of researchers to develop autonomous driving vehicles for urban environments. Similarly, the UK has committed nearly $25 million to support teams participating in the challenge.

The challenge is to create a system with a high degree of autonomy that can detect, identify, monitor and report a comprehensive range of military threats in an urban environment. The Grand Challenge is open to the whole UK science and technology base, large and small companies, research laboratories and academic science faculties. It is also a challenge to UK financial institutions to invest in UK. The Ministry of Defense (MOD) is keen to see the best solutions developed quickly into equipment for the UK Armed Forces.

You can read a more verbose description of the Grand Challenge here. Compared to DARPA's Urban Challenge, there are no restrictions on the number and type of robots that can be used. For example, contestants might choose to use a team of flying robots, ground robots or a mixture of both types. The MOD also expects the systems developed to be small (easy for a soldier to carry) and inexpensive (cheap to replace.)

There is no cash prize for the winner, instead,"the winner will have the kudos from their peers of being the best."

The final event will take place in the summer of 2008. The location for the final has been selected to be a mock-up village specifically designed for training troops in urban combat (see aerial photo on the left.) The teams will not have access to the location with the exception of a single visit in January 2008.