Ravens are smarter than you think

I am sure we all have observed a raven exhibiting behavior that most would consider intelligent. But considering that intelligence is hard to measure, do we know how intelligent ravens really are? Bernd Heinrich and Thomas Bugnyar are two researchers who set out to measure how intelligent ravens really are. They discovered some very fascinating facts about the common raven (Corvus corax) and they have published an overview article in this month’s issue of Scientific American.

The researchers set out to find out if ravens are capable of problem solving. They setup an experiment that required birds to retrieve a treat that was hanging from a string on a branch. This is how the authors describe the task,

To get a treat suspended on a string tied to the perch, a raven has to follow a precise sequence of steps-reach down and grasp the string, pull up on it with enough pressure to hold it there, let go of the string, and repeat the process.

They repeated the experiment with birds of different ages and they discovered that the older, more experienced birds could solve this problem in about 30 seconds, a sign that they use logic for problem solving. Younger birds either could not solve the problem or it took them much longer involving a bit of trial and error. The researchers speculate that the ability to think logically is acquired with experience as the bird ages which would explain why the younger birds could not complete the task.

The authors speculate that the raven’s playlike behavior is what enables it to learn problem solving. The article also discusses experiments that reveal the raven’s ability to distinguish among individuals, a skill that is important to them when it comes to storing food and keeping it safe from others.

I know that after reading this article, I will be seeing ravens in a whole other way.

BAE Systems unveils the Talisman M underwater mine-clearance robot

Talisman M is a new Unmanned Underwater Vehicle (UUV) specifically designed for the clearance of mines underwater. BAE Systems announced the economy car sized robot a couple of days ago promising revolutionary abilities that would remove human operators from the loop. Currently, teams of divers or remotely controlled underwater vehicles are used for mine detection and clearance. Poor visibility underwater often renders these remote controlled robots useless while surface detection methods return too many false positives requiring a human operator to verify possible mines.

The Talisman M is a step forward to minimizing the presence of a human operator in mine clearance operations. Part of the task can now be autonomously executed by the UUV although permission to deploy mine clearance equipment still requires verification by a human. When it is time to destroy a mine, the Talisman M will use one of four BAE Archerfish, small one-shot ROVs with cameras and destruction warheads. The Register has an in depth article about the capabilities of the Talisman M which regardless of how revolutionary it is or not, it is another step forward to saving the lives of the people serving in the military.

Visually tracking a tennis ball in video

We have previously discussed work on tracking soccer and hockey players in broadcast video using computer vision algorithms. This requires the development of robust tracking algorithms. Work at the University of Surrey, U.K., is focused on developing such robust data association algorithms for tracking objects in cluttered environments. The group has adapted a layered approach that starts with identifying a large set of candidate hypothesis with a high probability of containing only true positives. Additional steps then refine this set of candidates to identify and robustly track the object of interest. The team has demonstrated their approach by tracking a tennis ball in broadcast video. They presented their work at the IEEE International Conference on Computer Vision and Pattern Recognition (CVPR 2006.) Below is a video from the team that explains how the tracking algorithm works with a demonstration on a live tennis game,

Dancing robots are always fun to watch

We are all familiar with QRIO’s well choreographed performances of Japanese traditional dancing and more. It would appear, however, that the now defunct Sony QRIO has some artistic competition from no other than the almighty WowWee robots. The good folks from Q4 Technologies that developed the Go-Robo robot programming environment have created a video showing three Robosapiens, one Robosapien V2 and one Roboraptor robots dancing to The Tokens’ classic hit, “The Lion Sleeps Tonight.” Watch the video,

PS: I was really busy the last few days and I could not update the blog. However, I am done with most of my work and I will be posting regularly from now on.

Robot produces drip paintings but is it art?

Action Jackson is a simple robot built by spare parts for a course project. Its designers are computer scientists at the Washington University in Saint-Louis (WUSTL) led by Professor William D. Smart. The robot’s name, Action Jackson, is a tribute to the famous drip painter Jackson Pollock who according to Wikipedia

was an influential American painter and a major force in the abstract expressionist movement. He died in a car accident near his home in Long Island, New York.

No. 5, 1948 is one of his abstract, drip paintings that sold for $140 million in 2006. Action Jackson is a simple robot that creates drip paintings similar to Polock’s but not as complex. Watch the following video of Action Jackson painting,

A St. Louis Post-Dispatch article looks into the issue of whether Action Jackson’s creation can be considered art or better if the robot can be thought of as an artist. The article claims that the robot’s creation can hardly be considered art contrary to Smart’s opinion, who says,

If Action Jackson is a paintbrush — an extension of the artist's hands — then the seed of creativity is in the program that controls the machine. That seed is, most often, a program called "404," named after the number of the class in which Action Jackson was built. [Mechanical engineering student Topher McFarland] programmed the nozzle to trace out those numerals, sloppily, across the cardboard.

Now, I am neither an artist or really claim to understand art but the issues surrounding a machine’s ability to create art are very complex. According to the American Heritage Dictionary, an artist is defined as

1. One, such as a painter, sculptor, or writer, who is able by virtue of imagination and talent or skill to create works of aesthetic value, especially in the fine arts.
2. A person whose work shows exceptional creative ability or skill

So, can we really claim that Action Jackson is creating work of aesthetic value “by virtue of imagination?” Can a robot imagine? Is executing an algorithm, i.e., a well defined sequence of steps, imagination?

I always thought that an artist creates art expressing his/her view about certain events or just as a visualization of his/her own emotional state. Is Action Jackson trying to communicate a bigger message via its paintings just like Pollock and every other artist do? Can the robot look back at its own art and explain what it means? I guess not, so I would not really call this robot an artist. It is a pretty cool course project though!

Hack turns Robosapien to Robocop

Marcus from Evosapien.com either has too much free time in his hands or really likes hacking his robots. Recently he uploaded instructions telling us how to hack Robosapien V2 such that it handles a homemade BB gun with laser sight controlled via the WowWee supplied remote. Marcus calls his creation the CoilOsapien. You probably have to have some experience with electronics before you set out to implement this hack. Also, if you build it, make sure not to aim at anything that can brake because the gun can sure fire those pellets at high speed. With hacks like this, they can definitely revive the now defunct Robot Wars TV show. Why waste you time watching some remote controlled cars with chainsaws attacking each other when you can have a bunch of Robosapien robots shooting each other’s heads off? Add a few FlameOsapien robots (another of Robosapien hack by Marcus) in the mix and you will have one heck of a spectacle. I hope the Korean Robot Ethics Charter includes a clause about home hacks to turn service robots into killing machines (just kidding!) Here is a video Marcus made that demonstrates CoilOsapien,

John Backus passes away at 82-years young

The man who pioneered computer programming when he developed the FORTRAN programming language in the 1950s while working at IBM passed away on March 17th. FORTRAN was the first ever high-level programming language. Backus was also the developer of the Backus-Naur Form (BNF) notation for specifying context-free programming languages. He was also the recipient of the very prestigious ACM Turing Award in 1977. There is little else that I can write here that is not already said in articles published in the popular media.

Read more at Washington Post, Forbes, New York Times and the Inquirer.

Roboethics workshop at ICRA 2007

Robot ethics is quickly gaining ground as service and military robots are slowly entering our daily lives. The Institute of Electrical and Electronics Engineers (IEEE) in association with the Robotics and Automation Society (RAS) has established the Roboethics technical committee with the aim “to provide the IEEE-RAS with a framework for analyzing the ethical implications of robotics research, by promoting the discussion among researchers, philosophers, ethicists, and manufacturers, but also by supporting the establishment of shared tools for managing ethical issues in this context.”

Roboetchis will hold a technical workshop this coming April during the International Conference in Robotics and Automation (ICRA) to be held in Rome, Italy. The goal of the workshop, as stated on the official website, is

a cross-cultural update for engineering scientists who wish to monitor the medium and long effects of applied robotics technologies

The one day workshop includes a number of talks on the issues around Human-Robot Interaction (HRI,) the methodology of developing service robots and finally the implications surrounding the development of robot technologies and the actions undertaken by research scientists to address the ethical implications of such technologies.

Automatic tracking of hockey players in broadcast video

The automatic and simultaneous tracking of multiple non-rigid objects in video is a very difficult task for computers; this is a problem tackled by computer vision researchers and a good solution can have a large impact in developing intelligent systems including robots. Consider for example the problem of a robot tracking multiple people as it navigates in an office-like environment or when it is playing a soccer game (don’t forget that Robocup’s goal is to develop a team of soccer playing robots that can compete against the World Champions by 2050.)

Researchers at the University of British Columbia in Vancouver, Canada, are tackling the problem of tracking multiple non-rigid targets using an approach that they call the Boosted Particle Filter (BPF.) I will not go into details of how the BPF works but essentially it is a mixture particle filter with a proposal distribution derived from Adaboost. The people involved with this project are Kenji Okuma, Nando de Freitas and James Little. The BPF algorithm can be easily adapted for tracking any type of target but the team has demonstrated its effectiveness on tracking hockey players in broadcast video. Earlier, I had mentioned ASPOGAMO that is another approach to multiple hypothesis tracking applied to soccer.

Tracking multiple targets is a hard problem for many reasons. First, the fact that the targets are non-rigid means that their appearance in the video stream might change over time making it difficult to identify the same target over multiple frames. Second, targets may enter and exit the camera’s field of view at any moment meaning that the tracker must be able to instantiate new tracks while removing old ones. Third, broadcast video is hard to work with because the camera moves continuously to follow the play including zooming in and out of the action. Not knowing the internal parameters for the camera only further complicates the computer vision problem. Finally, the trajectories of the objects tracked could always cross paths making difficult to distinguish among the different targets. An example of this would be two hockey players involved in a collision.

The Boosted Particle Filter is a state-of-the-art approach to multiple target tracking that can handle many of the above issues. The following video shows tracking of a single player,

There are more videos and information about this project at Okuma’s website. There, you can also find an open source implementation of the BPF for MATLAB. Finally, the paper that was presented at the European Conference on Computer Vision (ECCV) 2004, was the recipient of the best paper in Cognitive Vision award.

Microsoft working on simulated Robot Soccer Challenge

In its efforts to market their relatively recent Robotics Studio, Microsoft has developed a soccer simulator in hopes that it can be used in RoboCup 2007. Microsoft Robotics Studio uses the AGEIA PhysX Technology to create simulated but realistic environments for robotics research while making it easy to port the resulting controller to a real robot. Bill Gates recently wrote an article for Scientific American providing a brief overview of robotics history while making sure to promote their new product.

A couple of days ago, Microsoft announced their intention to sponsor a competitive demonstration of their physics-based 3D engine applied to robot soccer in RoboCup 2007. Currently, there is an open call for participants and interested parties can find information at the U.S. RoboCup website.

It is expected that initially teams will work with an early version of the simulator that only supports wheeled robots; this should give developers a chance to become familiar with the software and also work on team strategy. Microsoft has promised to develop support for legged robots by May.

Finally, Microsoft is not limiting itself to just simulation. They hope that in the future they will be able to work closely with robot manufacturers to develop real soccer playing robots with software development centered on the Microsoft Robotics Studio.

It is clear that Microsoft is really serious about making an impact in robotics over the next few years. They have already gained support from a number of robot manufacturers including Kuka, iRobot, SRI International, Lego Systems, Yujin Robotics, Coroware, Parallax, Robosoft etc. I suspect that they will continue to gather support both from the industry but also from academia and even hobby roboticists on the virtue that their software is free and well supported and documented. The Microsoft Robotics Studio has numerous tutorials for beginner and advanced users not to mention 10 hours of video tutorials. In fact, I may go ahead and download a copy myself.

With computer vision by your side you will never have to ask for directions again

MIT’s Technology Review recently published an article describing how Augmented Reality (AR) will have a large impact in the way that we find information; they consider on of the 10 emerging technologies in 2007. Augmented Reality is a specialization of computer vision that attempts to extract the 3D structure of a scene from video and then insert with high precision virtual or 3D objects into the scene. AR can be useful in many applications including entertainment, heads-up displays, tourism and surgical visualization.

Consider for example visiting a foreign country and not having a map. If you want directions to your hotel from any location in the city then you can just take a photo with your cell phone camera and using the phone’s build-in AR software determine your location; the AR software would then retrieve directions wirelessly from a large database and present them to you by annotating the original photo. Another application would be retrieving information about a historical monument that you are looking at. An example of AR is shown in the following video from related research at the University of British Columbia,

Nokia is currently in the early stages of testing a cell phone equipped with AR algorithms to identify and provide information about a city’s landmark buildings. According to the Technology Review,

Last October, a team led by Markus Kähäri unveiled a proto­type of the system at the International Symposium on Mixed and Augmented Reality. The team added a GPS sensor, a compass, and accelerometers to a Nokia smart phone. Using data from these sensors, the phone can calculate the location of just about any object its camera is aimed at. Each time the phone changes location, it retrieves the names and geographical coordinates of nearby landmarks from an external database. The user can then download additional information about a chosen location from the Web--say, the names of businesses in the Empire State Building, the cost of visiting the building's observatories, or hours and menus for its five eateries.

You can find more examples of Augmented Reality at Nokia’s Mobile Augmented Reality Applications (MARA) project site and also at David Lowe’s website at the University of British Columbia.

Yujin Robotics line of entertainment, service and military robots

It is well known that Korea is one of the most robot friendly countries on the planet. There are numerous new robots introduced into the market every year ranging from remote controlled toys to military semi-autonomous mobile robots. Yujin Robotics is another Korean company that brings to market a large range of mobile robots.

The iRobi is a small entertainments robot with an expressive digital face and a large LCD screen on its chest for user interaction. The 45cm tall, wheeled robot can function as a mobile organizer keeping track of TV schedules and as an educational toy reading stories to young children. It also dances and it can take digital photos with its on-board 1.3 Megapixel camera. The iRobi can also function as a guard or remote surveillance device. In addition to the iRobi, Yujin Robotics also manufactures and sells soccer playing robots that can be used for the small robot league in RoboCup.

The company also is marketing an iRobot Roomba robot vacuum cleaner called the iClebo. The robot promises to do Roomba’s job with minimum noise.

Military robots that Yujin Robotics has developed include the Robhaz and Transbot. The Robhaz looks like a clone of the iRobot Packbot while the Transbot is essentially a real-life transformer robot. The Transbot can reconfigure itself to be either a two-legged walking robot or a four-wheeled car. The Transbot can be configured to carry a rifle, machine gun or rocket launcher and it can be controlled remotely using a Playstation-type controller. Here is a video of the Transbot in action,

Solving the localization problem by studying the rat’s hippocampus

Neuroscientists at the Department of Physiology and Pharmacology, Robert F. Furchgott Center for Neural and Behavioral Science, State University of New York Downstate Medical Center, New York, U.S., are studying how rats interpret spatial information by analyzing the activity in the hippocampus.

Specifically, Dr. Andre F. Fenton has published a recent article in the latest issue of Science discussing the latest discoveries about how different cells in the rat’s hippocampus behave allowing the rodent to identify previously visited areas. For example, activity in specialized neurons called “place cells” reveals that some discharge rapidly when the animal is in a part of the environment that it has visited before. Scientists observed the activity in the hippocampus when the animals were placed in different boxes with shapes varying from a square to a circle.

Knowing a rat’s location from the activity of its neurons is astonishing given that rats, like people, have no specific spatial sense organs analogous to, for example, the visual or auditory systems. Somehow, spatial knowledge is assembled by the brain.

In the past, researchers at the University of Queensland have proposed RatSLAM that is a solution to the Simultaneous Localization and Mapping problem in robotics inspired by the functionality of the rat’s hippocampus. The RatSLAM algorithm constructs topological maps that is useful for navigation. The team has demonstrated the visibility of their method for a visually guided mobile robot exploring an office-like environment without human supervision. There is more information on RatSLAM at Michael John Milford’s website; Milford is the primary researcher on RatSLAM. You can also watch a video of a robot navigating using RatSLAM here.

The Semantic Robot Vision Challenge

For the first time, researchers will compete in the Semantic Robot Vision Challenge (SRVC) as part of AAAI’s Mobile Robot Competition and Exhibition. The idea behind SRVC is to test machine learning algorithms applied to computer vision using online databases. The competition will essentially send mobile robots to a scavenger hunt.

The Semantic Robot Vision Challenge is a new research competition that is designed to push the state of the art in image understanding and automatic acquisition of knowledge from large unstructured databases of images (such as those generally found on the web).

Each robot will be given a textual list of objects that it must identify by searching in a confined area. It will first be allowed a small amount of time to go online and search on Google images for example photos of what the objects look like. The robot should use this information to learn a model of each object’s appearance without supervision. Lastly, it will be allowed to enter a given area and autonomously search for the objects by using its on-board camera. The winner is the robot that identifies the most objects correctly.

This challenge is not an easy task. The machine learning part is tricky because the robot must work with a large collection of images to be returned by Google. In addition, many of the example images will have lots of clutter making it difficult to identify the visual characteristics of just the object searched for. For example, if the robot is searching for images of bicycle helmets then many of the results returned by Google will likely include photos of people wearing helmets making it difficult to segment just the helmet without human supervision. The robotics part will also be difficult because it will require autonomous navigation and planning in a very large state space.

The final set of rules will be posted in just a few days on March 15th. The Semantic Robot Vision Challenge will take place during the 22nd National Conference on Artificial Intelligence to be held July 22-25 in Vancouver, Canada. The SRVC is also sponsored by the National Science Foundation.

I am really curious to see what kind of solutions people will come up with in order to complete this object discovery task.

NASA robot to repair and refuel satellites

NASA in cooperation with the Defense Advanced Research Projects Agency (DARPA) has developed and launched into space a robot that will take on the difficult job of repairing and refueling military and civilian satellites. Orbital Express is a two spacecraft system that for the next three months will be used to demonstrate the feasibility of having robots service orbiting satellites.

The repair robot is named ASTRO (Autonomous Space Transport Robotic Operations) while the satellite used as a dummy for repair and refueling mission is called NextSat. ASTRO is equipped with a small robotic arm that it can use to service NextSat. The two vehicles are worth a combined $300 million U.S. dollars.

NASA hopes to use these robots to extent the lifetime of satellites which now are only good for one use. When the satellite runs out of fuel, it is destroyed using a controlled reentry. NASA currently uses the space shuttle to repair satellites if necessary. The shuttle usually captures a satellite using its robotic arm and brings the vehicle in its bay where astronauts perform the repairs. Since the shuttle will be retired in a few years and a replacement vehicle is still in the works, it makes sense that NASA is seeking an alternative for servicing satellites.

It is unfortunate that Orbital Express will not be available for deployment for another 10 years or so. The Hubble Space Telescope is in need of servicing. In the past a shuttle mission could be dedicated to the task but since the unfortunate destruction of Columbia this is no longer feasible; except for one last mission in late 2008. Researchers had proposed developing a robot to repair Hubble but unfortunately none was able to develop one in time to save the telescope.

In robotics, you can have your cake and eat it too

Every other week, I blog about either the development of some new gun totting robot or the efforts of law makers and scientists to establish ethical guidelines for robots. The South Koreans were the first to give a machine gun to an autonomous robot with the announcement of the SGR-A1. The U.S. based iRobot continues to develop the Packbot robotic platform to help in policing and combat while the British and Korean governments are moving forward in developing ethics codes for robot development and use. Today, Elbit Systems Ltd. announced the creation of another robot designed for combat use.

Elbit’s VIPeR remote controlled platform is small and lightweight enough to be carried by infantry in the battle field. The 25-pound, 9-inch tall wheeled robot is capable of climbing stairs, navigate autonomously and carry a mini-Uzi automatic pistol, stun and smoke grenades, explosives sniffer and day and night vision cameras.

There is no more information about VIPeR that I can find other than the one photo shown above. We know that the robot is not yet in production and that Elbit plans to begin field trials with the cooperation of Israel’s military before the robot is deployed along with infantry units.

For the record, the VIPeR is the second robot to be specifically designed to carry weapons although compared to the SGR-A1, it cannot fire them on its own but only under the instruction of a remote control operator.

PaPeRo to become your personal blog assistant

PaPeRo is NEC’s cute, little robot designed for assisting with information discovery and entertainment. The original robot was designed to help its owner retrieve information about daily tasks such as TV schedules and weather forecasts. Additionally, PaPeRo is an entertainment device armed with a variety of skills including story telling, singing, dancing or just acting wacky.

Recently, NEC announced an upgrade for PaPeRo that will essentially turn it into your blogging personal assistant. Armed with state of the art natural language algorithms and access to the Internet, PaPeRo can record video and audio of its user detailing a story that could be posted on his personal blog. PaPeRo analyzes the spoken text and searches online for related multi-media content that can be added to the blog augmenting the original story.

This system of course can be very useful if the Natural Language Processing (NLP) and speech recognition software works well. I have not seen the system in action but I would guess that the robot’s owner would have the last word when it comes to which material discovered online get to be posted on the blog; there is little chance that PaPeRo will be able to find only the relevant and best information considering the amount of spam sites out there. Even Google has a hard time filtering out all the garbage sites.

That said, I think this would be a cool application except for one problem. What is the advantage of having a robot to assist with blogging? NEC could easily have developed a software agent with a virtual body that could interact with the user. I understand that there exists a virtual version of PaPeRo and maybe that is the intended application for the assistant functionality and so the use of the actual robot version is just for good PR.

Lastly, it is very interesting that companies are starting to realize that there is lots of information online that can easily be tapped to help improve intelligent systems. In fact, AAAI will be running a new intelligent systems competition centered on the same idea. But I will write about that tomorrow.

Robosapien becomes a movie star

When I first saw the press release, I thought that it was an April fool’s joke but then I checked the date and it was still the first week of March. It would appear that there is no stopping the WowWee train. The robotics company that has impressed many with their diverse line of robotic toys has announced that they are joining forces with Arad Productions in order to create a movie with Robosapien as the main character!

The press release gives little information about the movie’s plot and only says that it will be a mix of live-action and CGI. The movie will be released in 2009.

The producer will be no other than Avi Arad who is the person behind a number of recent movies based on comic book characters. Some of his credits involve Spider-Man 2/3, X-Men: The Last Stand, Ghost Rider, the Hulk, Daredevil and many others.

Once again, I have to admit that I am totally impressed with WowWee's marketing genius. WowWee appeared out of nowhere and in a couple of years this privately owned company has become associated with entertainment robots. They recently completed a deal with McDonalds that has brought miniature versions of their robots to the restaurant’s Happy Meals. WowWee has sold more than 4 million robots to date and I am sure that they will sell many more in the years to come.

Remote controlled robot to help solve the mystery of Earth’s missing crust

A 12-member team of British scientists are on the way to the middle of the Atlantic Ocean in an effort to answer the questions around Earth’s missing crust deep in the bottom of the ocean. According to CNN,

The hole is about 16,400 feet under the surface of the Atlantic and located half way between Tenerife and Barbados. It has a diameter of 10,000 to 13,000 feet.

The team will use a remote controlled underwater vehicle to try and recover samples from what should be the Earth’s exposed mantle; mantle is the material that is found under the Earth’s crust but cannot be reached normally.

Because of the extreme depth, a manned submarine mission is not possible. As a result, the team will use an autonomous underwater vehicle that is essential a remotely controlled robot with enough on-board decision making to make the operators job easier. The robot will land at the bottom of the crater and drill into the mantle in order to return rock samples back to the ship. In addition, the robot will use its on-board mapping instruments to create a 3D map of the ocean’s floor.

You can follow the team’s progress online at the Classroom@Sea Project website.

South Korea's Robotics Ethics Charter

The British were the first to seriously consider the ethics in living with robot last December when the government sponsored a speculative paper suggesting that robots one day might demand equal rights to humans. Today, the South Korean Ministry of Commerce has announced that a team of 5 people including scientists, futurists and a science fiction writer will come together to draft the Robotics Ethics Charter.

The document is meant to be a guide about how people should treat robots and vice versa. Apparently, the team of experts will be basing the Charter on Asimov’s well known three laws of robotics. The South Korean government had announced last year that they plan to have one robot in every home in the next 10 to 15 years. Most of these robots will be used as servants to care for an ageing population in a country with low birth rates and faced with a shortage of qualified workers.

First of all, I believe that the fact people are talking about robot ethics many years before living with robots becomes a reality is the right thing to do. We should be ready for when robots are everywhere so that we don’t have the same problems such as those created with the sudden growth of the Internet in terms of free speech and privacy.

Second, I believe that any document based on Asimov’s three laws of robotics is doomed to failure. Azimov introduced these basic laws in a short story and then wrote much about how these laws were not perfect, conflicting and could easily be misinterpreted by the robots. Many people have modified these laws since then and in fact even Azimov eventually added a zeroth law.

Finally, I find it very ironic that the people mostly concerned with robot ethics and in fact are supporting Azimov’s three laws of robotics are the same people who have created the SGR-A1 military robot. If South Korea moves forward with their plan to actually deploy this robot along its border then how exactly are they applying the first law of robotics that reads,

A robot may not injure a human being or, through inaction, allow a human being to come to harm.

AAAI changes its name to reflect its international audience

The American Association for Artificial Intelligence (AAAI) has changed its name to the Association for the Advancement of Artificial Intelligence (still AAAI.) The reason, as given by AAAI president professor Alan K. Mackworth, is based on the fact that a large portion of AAAI’s membership is no longer limited to the U.S.

In fact, many AAAI sponsored conferences and workshops are held outside the U.S. and many of the paper submissions to these events also are from an international audience. As Mackworth puts it in a letter published on the association’s website,

AAAI’s membership has strong international representation. The same is true of the contributors to, and attendees of, AAAI, and AAAI-sponsored, conferences, symposia, tutorials, and workshops. A large fraction of our membership and an even larger fraction of conference paper submissions are from authors based outside the U.S. Beyond the annual conferences held in North America, much of AAAI’s mission, activities, and membership are not well-characterized by the label “American.” In short, our original name does not reflect our dual roles, national and international, in the global AI community.

AAAI was founded in 1979. The main, annual AAAI conference will be held in Vancouver, BC, Canada, this summer. AAAI is also the organizer of a number of AI and robot competitions. They also publish AI Magazine and sponsor a large number of workshops and symposia.

AMD announces its own teraflop capable system to rival Intel

Not to be outdone by competitor Intel, AMD announced a couple of days ago their own teraflop in a box system. Last month, Intel unveiled their plans for an 80-core CPU that could perform 1 trillion floating point calculations per second; Intel’s Teraflop Research Chip is not expected to be available for another 5 years and most likely it will be targeted to the server market at first.

AMD, in the meantime, is taking full advantage of their recent acquisition of graphics accelerator manufacturer ATI. Just as March rolled in, AMD announced their approach to developing a Teraflop computer. Intel is trying to get the increased performance with a unique CPU design that packs many cores; AMD is outfitting a dual core Opteron processor with two R600 graphics accelerator cards to achieve extraordinary performance in a single box but not a single CPU. AMD’s Accelerated Computing platform is a single system that should be available in the next 1-1.5 years much sooner than Intel’s Teraflop Research Chip.

The Accelerated Computing platform will initially benefit scientific, medical and business applications. One aspect of the new design that AMD has not made clear is the power requirements for the new system. Intel’s research chip promises not only high performance but also low power consumption of less than 100W. One can speculate that AMD’s system might require anywhere from 100W to 300W (or more) considering the power requirements for the Opteron and R600.

During this past year, there has been much progress in developing co-processors for specialized floating point calculations mostly driven an interest for general scientific computing on GPUs. For example, ATI had recently announced their support for Stream Computing and NVIDIA released specialized C, C++ compilers for easily developing non graphics-related applications on their graphics accelerator cards.

Autonomous exploration with a vision-based mobile robot

In recent years, the vast majority of mobile robots have been constructed to rely heavily on a laser sensor for perceiving their surrounding environment. Laser is very accurate when it comes to measuring the distance to objects making it possible for mobile robots to navigate and map large and highly dynamic environments. On the other hand, laser sensors can only scan along a single plane and often fail to perceive large objects such as tables; a laser sensor would only detect the legs of the table.

Stereo vision is a much better sensor for use with mobile robots because not only can it measure distance to objects but image data can also be used for other tasks such as localization, object and people recognition. Researchers have avoided using stereo vision on their mobile robots because in the past computers were not fast enough to process all the data and also because we did not have the basic algorithms to make computer vision work.

Recent advances in computer vision have made it possible to construct visually guided mobile robots that are as capable as the robots depending on laser. For example, Rob Sim and his colleagues at the University of British Columbia, Canada, have developed a mobile robot that can estimate its location and construct a map of its environment using only visual feedback.

Last summer, Rob demonstrated the capabilities of his robot when he published a paper showing it autonomously exploring a large office-like environment correctly estimating its position using a Rao-Blackwellised particle filter (RBPF) and in addition constructing an occupancy grid map. An occupancy grid is a 2D plan of the environment such that the area occupied by stationary objects is marked along with the space that is free and can be traversed by the robot.

Here, I have included a video that shows the map construction process. The robot travels more than 100 meters mapping an area more larger than 400 square meters. The robot is shown with a yellow rectangle while the path it follows is in yellow. The red path displays the direction that the robot decides to move in order to maximize the accuracy of the learned map and in blue is the raw odometry or better the robot's uncorrected trajectory. Black pixels corresponding to occupied space, i.e., walls and furniture, while white pixels denote empty space. Grey pixels denote space that the robot has not observed because for example it is outside the camera’s field of view or it is space hidden behind another object. There are more movies showing the mapping algorithms in action here.

Anybots dynamically balanced walking humanoid robot

A small technology company in California announced this week that they have built the first dynamically balanced walking humanoid robot. The company, Anybots, released a video of the robot Dexter walking in a way similar to the way humans walk. The robot essentially is continuously falling forward using its legs to break its fall and essentially walk. Dexter is 5 feet 10 inches tall and weighs 135 pounds although it has no arms.

Dexter’s walking abilities are definitely extraordinary but the company’s claims that this is the first dynamically balanced robot are a bit exaggerated. All legged robots are dynamically balanced including for example Honda’s ASIMO. They just happen to use a different balancing algorithm and mechanics. In fact, one of Dexter’s most interesting aspects is the use of pneumatic actuators instead of electric motors used in most humanoid robots. Dexter is also claimed to learn how to walk using machine learning which is no small achievement in itself.

In addition, Anybots is a small company of only 3 people and they have been developing this robot since 2001. Dexter is an extraordinary achievement for such a small team which is also developing another robot, Monty. Monty has an articulated arm driver by 18 motor while it navigates using wheels dynamically balancing similar to a Segway.

Here is the video of Dexter walking,