Bioloid robot kit

If you are looking for a cool robot kit for Christmas and have a few hundred dollars to spend, then you should check out the Bioloid robot kits and especially the expert package. The beginner kit is more suitable for those new to the world of robotics including a simpler programming interface and easier robots to assemble. The expert kit is a step up with 26 robot examples programmable using C in Microsoft Windows. Using this kit, you can construct a number of different robot including simple wheeled robots, a robot dinosaur, a robot dog, a six-legged spider robot and a humanoid robot. The cost for the expert kit is $3500 dollars but if you ask me it is totally worth; you can also purchase the more affordable comprehensive kit (for building a humanoid robot) for $899. Check out the photos below (click for larger images) of the kits and the collection of robot you can build with it.




The kit is available at Robotics.com.

On my way to the International Conference on Intelligent Robots and Systems

Next week, I will be attending the 20th International Conference on Intelligent Robots and Systems (IROS) to present some of my recent work on robotics. The conference which will take place in San Diego, California, was in danger of being canceled because of the wildfires that ravaged much of Southern California this and last week. Thankfully, the fires are now under control and the conference organizers who kept us informed almost daily with up to date information have announced that the conference will take place as originally planned.

IROS is the second largest robotics and automation conference The largest one is the International Conference on Robotics and Automation (ICRA) while the third largest is Robotics: Science and Systems (RSS); the latter is a relatively young, single track conference that is slowly gaining momentum. To give you an idea of how large IROS is, consider that more than 1000 researchers attend every year to present more than 400 papers; the conference has 11 simultaneous tracks (that is 11 presentations going on at the same time) and 2 days of workshops and tutorials. There is also an exhibition and a number of plenary (invited speaker) talks. And finally, just like all major conferences, organizers will present several awards for best papers and excellence in robotics research.

I am looking forward to the conference; I will try my best to post photos and videos of all the happenings at IROS. Stay tuned!

Evolution Robotics and Wowwee join forces

Evolution Robotics makers of robotics software and Wowwee makers of robotics hardware (toy and entertainment robots) announced today a partnership to deliver jointly a series of new robotic entertainment products a bit more than a year from now. The companies expect to release 3 new robots in 2009. Let me remind you that Evolution Robotics develops software for robot programming and image processing including object recognition. Wowwee has found much success with their toy robots since the introduction of the first Robosapien humanoid in 2004. Evolution Robotics image processing software was also used in the now defunct Sony AIBO robot dogs; it was their software that allowed AIBO to recognize the different command cards.

According to the Press Release,

Each product will be powered by Evolution Robotics’ latest NorthStar 2.0 technology and deliver breakthrough robotic capabilities that have never before been available to the robot and robotic toy mass market.

"Evolution’s technologies are revolutionizing the way we interact with robots and consumers will be truly amazed when they see what these products can do," said Davin Sufer, Chief Technical Officer of Wowwee. "Together we plan to create an entirely new category of products to match the excitement and interactivity of video games, with blazing speeds and real-time action."

The product initiatives are the result of a strategic partnership between the two companies involving 18 months of research and development to design an innovative positioning and tracking system that will enable a robot to autonomously interact with people, environments, other robots and objects with pinpoint accuracy and control.

Under the latest agreement, Wowwee and Evolution Robotics will work closely together to take the completed products to market, leveraging the strengths and skills of each company. Wowwee will manage the product manufacturing, marketing and sales, and Evolution Robotics will provide the key enabling technologies and apply their expertise in robotics to integrate the products’ functions.

This looks like a very exciting partnership which spells good news for robotics enthusiasts. I'm looking forward to the first few products resulting from this partnership.

Reinforcement Learning is cool

The following video was created by the Real-life Reinforcement Learning group at Rutgers University and, in my opinion, it demonstrates why reinforcement learning is awesome! Using RL, artificial intelligent agents such as the AIBO shown in the video can learn solutions to problems that a human engineer may not consider herself. Sometimes the solutions are very creative such as the one found by the AIBO in the video below.

The video won the award for Best Short at the first AAAI video awards last summer.

If you want to learn more about Reinforcement Learning, then take a look at the introductory RL book by Richard Sutton which is available for free online here.

Artery cleaning micro-robot

The Telegraph is reporting on a new medical micro-robot from Korea that is small enough to enter and navigate the arteries of a human being, carry medication, and use it to unclog arteries preventing heart attacks. The robot is tiny measuring less than a millimeter in length and boasts 6 legs. One of the most interesting aspects of this micro-robot is the power and locomotion systems.

By attaching grafted heart muscle to the legs, the scientists found the legs would bend as the muscle cells contracted. The cells get their energy from sugar in the patient's blood.

That means the robot does not need an external power supply, which are often heavy and cumbersome, if not impractical.

Because the robot's three front legs are shorter than the back legs, they bend inwards as the heart muscles contract, creating a difference in friction that pushes the robot forward.

This is damn cool and I want to know more about it but the article gives no information about the people behind this project; it only says that the work originates at the Chonnam National University. I could not find any information about the project on the school's website (at least the English version.) If anyone knows more about this project, let us know in the comments.

Digital camera image rendering directly to graphics card

Point Grey Research has released a new version of the Ladybug API with new features that allow the camera to directly render images in the computer's graphics card memory; these images are then easily accessible as textures. Ladybug is the company's spherical vision product; see an example of the images captured using a Ladybug camera mounted on the roof of a moving car.

The updated library now uses hardware acceleration to rectify captured images for a 3X speedup compared to the non-accelerated version. New functions added to the API allow for easy access of image data stored on the graphics card. Since the images are stored as textures, they can easily be used to texture map any geometric object in a 3D environment. The new SDK is loaded with examples demonstrating this new functionality. It won't be long before more image processing becomes mainstream on graphics cards including color calibration, stereo, optical flow, and possibly image feature detection.

Mars rovers get 2 year mission extension

NASA has approved a 2-year extension for the most successful pair of robotic rovers to ever land on the Martian surface. The two rovers, Spirit and Opportunity, were initially designed for a 3-month mission which has repeatedly been extended because of the robustness and effectiveness of these planetary explorers.

The two robots are the best example of great engineering and I am glad that NASA is not giving up on them. In case you are wondering, the odds of a spacecraft successfully reaching Mars let alone landing on its surface are very low as you can tell from the large number of unsuccessful attempts in past years by NASA and the European Union. In other words, the cost of $20 million dollars for the continued operation of the two rovers is nothing compared to the cost of building new robots to replace them; not that NASA and other space agencies are not readying new missions for Mars but why not get the most out of the current robots on the planet's surface? The original cost for the two rovers was $820 million.

Reinforcement learning competition

The machine learning community will be holding their second annual reinforcement learning (RL) competition designed to test new RL algorithms in a variety of complex domains including a real-time strategy game, soccer competition, Tetris, helicopter simulation and others. Specifically, the competition will include the following events,

Mountain Car: Perhaps the most well-known reinforcement learning benchmark task, in which an agent must learn how to drive an underpowered car up a steep mountain road.

Tetris: The hugely popular video game, in which four-block shapes must be manipulated to form complete lines when they fall.

Helicopter Hovering: A simulator, based on the work of Andrew Ng and collaborators, which requires an agent to learn to control a hovering helicopter.

Keepaway: A challenging task, based on the RoboCup soccer simulator, that requires a team of three robots to maintain possession the ball while two other robots attempt to steal it.

Real-Time Strategy: A game, based on popular real-time strategy games, which poses exciting new challenges for the reinforcement learning community.

Polyathlon: The agent will face a set of potentially unrelated MDPs with minimal task knowledge and no prior training.

The organizers will release the competition software in 2 weeks time while they will begin accepting team results as early as December of 2007. The final results are expected by July 1st, 2008 just in time for the final event during the International Conference on Machine Learning to be held in Helsinki, Finland, July 6-9, 2008.

More information at the reinforcement learning competition website here.

Robot sex?

David Levy , a recent PhD graduate from the University of Maastricht in the Netherlands, is telling the world that in the not so distant future humans and machines will fall in love, have sexual intercourse and marry. Levy recently outlined his theory of "interspecies" (if you can call it that really) marriage in his thesis Intimate Relationships with Artificial Partners.

There is already a trend towards people developing emotional bonds with their household robots as this article about the connection between Roomba vacuum cleaning robots and their owners suggests. Levy considers the fact that robots will eventually be designed to look and act more like humans and possibly be given certain rights as citizens. As a result, it is inevitable that at some point in the future, a machine and a human will be able to legally marry; this would be something with larger consequences than the machines just used as devices for sexual relief. Lots of scientists argue that even though sex with a robot is inevitable, legal marriage maybe far fetched. They don't believe that these robots will function as nothing more than a glorified sex toy.

At any rate, Live Science has a more complete article addressing many of the issues surrounding the love affair between man and machine; you can read it here. In addition, I especially enjoyed reading Lewis Page's take on the subject as he wrote it for The Register and published here.

On the more amusing side of things, I suggest you watch the Futurama episode titled "I dated a robot" for an interesting and hilarious look at what our future with dating robots may be like; I mostly like their "Don't date robots" educational clip which unfortunately I can't find by itself anywhere online.

iCub: The open source humanoid project

A European consortium is working full-time on a new open-source humanoid robot known as iCub. When finished, iCub will have 53 degrees of freedom, stand 104 cm tall and weigh 23 Kg. This new robot will have twice the degrees-of-freedom compared to Honda's ASIMO and Fujitsu's HOAP-2.

At 104 cm tall, the iCub is approximately the size of a three year-old child. It will be able to crawl on all fours and sit up, its hands will allow dexterous manipulation, and its head and eyes are fully articulated.It has visual, vestibular, auditory, and haptic sensory capabilities.

Earlier, I said that this is an open-source robot because its creators will be making the complete specification available for free on the Robot Cub project website. In a way, the iCub project may be able to develop the open source humanoid robot that PINO failed to deliver.

As an open system, the design and documentation of all hardware and software is licensed under the Free Software Foundation GNU licenses so that the system can be freely replicated and customized.

Other than creating the most advanced and open humanoid platform today, the Robot Cub team specifically designed the iCub for studying human cognition. The team recently announced that they are accepting proposals for research in embodied artificial cognitive systems using the iCub platform.

An iCub kit will be awarded each of the applicants who submit the eight best proposals. In addition, each successful applicant will be awarded travel and accommodation costs at the Italian Institute of Technology in Genoa for one person for a two-month period to participate in the supervised assembly of the robot. Shipping costs of the assembled iCub to the applicant's host institute will also be covered.

The deadline for applications is December 1st, 2007. For more information, visit the Robot Cub website.



iCub photos are copyright the Robot Cub project.

Object recognition using a grid-connected robot dog

An interesting project at the University of Amsterdam, Netherlands, involves color-based object recognition using distributed computing consisting of several beowolf clusters spread around the world. The video at the end of this post demonstrates how the system works; the video is also the winner for "Most Innovative" at the first AAAI video competition held last July during the annual AAAI conference on Artificial Intelligence.

This demonstration shows object recognition performed by a Sony Aibo robot dog. The dog is connected to a wide-area Grid system, potentially consisting of hundreds of computers located at several institutes in Europe, the United States, and Australia. Apart from the quality of recognition, we demonstrate the effectiveness of Grid usage in multimedia computing. Moreover, we show the ease with which multimedia applications can be integrated with Grid computing.

Futurama movie: Bender's big score

I have to admit that I am a huge fan of Futurama and I was one of those seriously upset with Fox when they canceled the show. But luckily, Futurama is coming back on DVD in feature film format with all the original cast. In less than two months, Futurama is making a comeback with the all new movie titled Bender's Big Score. The plot, according to IMD is about,

Planet Express sees a hostile takeover and Bender falls into the hands of criminals where he is used to fulfill their schemes.

Sounds like it is going to be all Bender although the trailer shown below has a little bit of everyone, including my favorite character, Dr. Zoidberg. It probably won't take the criminals long to convince Bender to commit crimes since he is already a convicted felon. Enjoy the movie trailer,


This DVD is number one on my holiday wish list.

Age of Mythology game AI is worse than initially thought

Researchers from the Austrian Research Institute for Artificial Intelligence have now shown that even an irrational player can easily defeat the game AI in Real-Time Strategy (RTS) games. According to New Scientist, the team designed a number of “emotional” software agents playing different strategies ranging from defensive to neurotic and found that all outperformed the build-in AI for the RTS game Age of Mythology. In fact, the neurotic agent which often miscalculated the amount of available resources and acted irrationally won all games faster than all others.

Just to make it clear, the results say nothing about how well any of these intelligent agents will fair against a human player. But we can at least conclude that Microsoft needs to hire some smart people to develop better AI for their games. I understand that developing robust RTS AI is not easy (any large multi-agent problem is a hard problem) but if it can be defeated this easily then they surely are doing many things wrong. Considering the size of the computer game industry (apparently even larger than the movie industry in revenues,) I am continuously surprised that game publishers don't work more closely with universities to develop new algorithms for multi-agent, real-time AI.

MAARS: New transformer-like robot goes to war

QinetiQ North America announced today a new military robot that boasts a number of new hardware and software improvements over its predecessors. The new Modular Advanced Armed Robotic System (MAARS) military robot is designed to replace the TALON and SWORDS platforms. MAARS is a lighter robot that can move faster than the older models; it can also carry either a machine gun or a robotic arm so in a sense it is a single robot that can perform the tasks of the two other robots before it. When equipped with a robotic arm, MAARS becomes a perfect platform for IED removal.

The purpose-built MAARSTM chassis provides a uni-body frame with easier battery and electronics accessibility. Other features include a larger payload bay, higher torque, creating faster ground speeds and improved braking. The new Digital Control Unit significantly improves command and control and situational awareness for the operator resulting in greater safety margins. The complete system weighs about 350 pounds.

In terms of the software, Wired.com reports that the new robot can be configured with safe zones where it cannot shoot at people; these are meant to be areas occupied by friendly troops.

MAARS (Modular Advanced Armed Robotic System) features new software controls, which allow the robot's driver to select fire and no-fire zones. The idea is keep the robots from accidentally shooting a flesh-and-blood American. A mechanical range fan also keeps MAARS' gun pointed away from friendly positions.

This is an instance of controlled autonomy for mobile robots. In such a system, a human operator has control over the robot's actions at all times. A fully autonomous robot would not require any type of supervision. It makes sense that the military would want control over a machine that can pull the trigger at will. It is a mechanism that can prevent casualties among friendly troops and situations such as those portrayed in Terminator and The Matrix movies where machines decide to take over and eliminate human life.

The following video shows the MAARS robot.

Urban Challenge teams that stand out

CNet's News.com takes a special look at some of the upcoming Urban Challenge teams that stand out. In this special gallery, you will find information about the usual suspects from Stanford and CMU and newcomers from MIT and Europe including information about the technology and budget of each team. The month of October is very important for the participating teams because during the last week of the month, DARPA will hold the semifinals in Victorville, California at the former George Air Force Base. Some vehicles worth mentioning are Team Oshkosh's TerraMax truck and Georgia Tech's Porsche Cayenne which would be a shame if it was run over by the former!

Check it out at, "On the road to robot race."

Robotic swami conversational robot for under your Christmas tree

If you are looking for a special Christmas gift for a tech-oriented family member and you happen to have an extra $75K to spare then consider purchasing the robotic swami conversational robot. Driven by 30 micro-motors and AI software running on a laptop computer, the robotic swami promises to keep you, your family and guests entertained to the max.

I can't figure out who makes this robot, and I only found out about it via a CBC article reporting on the Neiman Marcus 2007 Christmas gift catalog for the elite. According to the product page at the Neiman Marcus website, the robotic swami is

A cutting-edge Artificial Intelligence robot that recognizes family members, can carry on conversations, develops relationships, and answers questions with fact-based advice, now that's going to freak out the kids. Say, "Hey" to the Swami.

The OMG factor on this dude is off the charts. You'll swear he's human, right down to his facial expressions. How does he do it? Revolutionary character-engine artificial intelligence software. Micro-camera eyes. More than 30 robotic micro motors. All running off a PC laptop (included, of course).

Teach him to wink every time the wife walks by. Name him Mortimer. Pack him up in his suitcase and take the show on the road.

I wish I knew what this revolutionary artificial intelligence software is; I'm guessing that it is probably some old-school rule-based system that can handle basic interactions to keep people interested in the device for a bit. I guess, considering its price tag, the company has to claim extraordinary capabilities or else who would really buy it? I'll just save my money for the more affordable i-SOBOT or (even) the much delayed Pleo.

Driftnet: Stereo vision art display

The Squidsoup crew, Anthony Rowe, Gaz Bushell and Ollie Bown, has put together a very interesting and entertaining display blending computer vision and art. The art display, Driftnet, gives people the ability to fly like a bird in a virtual environment. Driftnet depends on a Point Grey Research Bumblebee stereo camera to track a person and translate his/her motion to control commands (essentially like a whole body joystick.)

At one level, it experiments with intuitive methods for freely navigating 3D virtual space. Users are invited to ‘fly like a bird’ to navigate through a virtual space. Using NO worn equipment - just by flapping their arms/wings and tilting their arms and bodies – people can intuitively (and amusingly!) navigate freely in virtual space. The metaphor used harks back to childhood play, imitating birds and planes in the playground.

The camera is mounted on the ceiling looking down at the user. The image above shows the depth maps for different gestures; the different inputs that the system understands have been simplified to emphasize the artistic aspect of Driftnet.

The following video shows Driftnet in action,

JBoost: Free AdaBoost implementation

For those interested in machine learning and are looking for a nice implementation of a number of boosting algorithms including the basic AdaBoost and its variants LogitBoost, BrownBoost, BoosTexter and (soon) NormalBoost, you should check out the JBoost software from the University of California, San Diego. Other than the implementation of a large number of boosting algorithms, the software is optimized using Alternating Decision Trees (ADTrees) to significantly reduce the number of nodes in the learned classifier decision trees.