Take five: the PhD dilemma

I figured that I should take a break from reporting on robot news and cool projects at different labs around the world to talk about graduate studies. This post is basically inspired by Tony Darnell's recent post titled "The Best Reasons I know for NOT Going to Graduate School." Tony is an astronomer working full-time and considering a return to school to pursue a PhD. He is faced with the dilemma of going to graduate school or not and he is sharing his thoughts about the pros and cons of graduate studies. He outlines four reasons for not pursuing a PhD including, prestige, money, stress and what he calls pigeon-holing.

He says that doing a PhD for gaining prestige is not a good idea. I agree that you don’t need an advanced degree in order to be innovative and publish; in fact, anyone can submit a paper to a scientific journal and it will be reviewed with the rest of the papers not taking the author’s name and degrees into consideration. The truth is, however, that the chances of someone outside academia getting something published are slim to none. The reason? It takes a lot of time, conduct with academics working in the same area, graduate student labor and money in order to generate something valuable. I can’t imagine being able to work on writing papers and doing careful study of difficult problems if I have to work a 9-5 job and enjoy family life at the same time; unless of course, the 9-5 job is being a scientist.

I totally agree with him that money should not be your number one reason for getting a PhD but not because you won’t be paid well once you obtain your degree but because science is about the pursuit of knowledge and not about fattening up your bank account. It is true that during your degree you will make enough money just to get by but after graduation, you will be rewarded for your expertise. The problem is that the money you lost pursuing the degree when you could be working full-time cannot be recovered after graduation. I have estimated that during the 5 years that it took me to complete my degree, I forfeited nearly $250,000 in lost wages and I know that even with a six figure salary today, I cannot make that money back. So, yeah, don’t do a PhD if you are after money.

In terms of the stress level, I also agree with Tony. It is hell on Earth trying to complete your degree and after that try to get tenure. If you can’t handle stress very well, then steer away from graduate school and academia all together. Having a balanced life-style with a supporting family and friends can go a long way in terms of keeping your stress levels down and making your adventures in academia a more pleasant experience.

In terms of pigeon-holing, I have to disagree with Tony. He says that many students end up working on the research area of interest to their supervisor even though the work is of little to no interest to them. This adds to the stress levels since they are forced to work on something they don’t necessarily enjoy. Sure, this is true to a large extent. In my experience, however, most students entering a PhD program and having just finished their B.Sc. don’t have the necessary knowledge to select a research area for themselves. They really don’t and I see this every year. Supervisors have the hard job of trying to figure out what is of interest to a student and give them relevant work. That is what supervision is all about. In addition, students are often free to select a supervisor among the faculty in their department assuming of course that the professors have room for more students. To a large extent, it is up to the students to find a supervisor that works on what is of interest to them. Most new students entering graduate school practically select a supervisor blindly because they don’t know what is of interest to them in the first place or just because they don’t realize how important it is. Tony’s suggestion to new students that they should be “careful what you pick because you are likely to be stuck with it for the rest of your life” is more or less valid. On the other hand, if you really keep an open mind and are curious enough to explore more subjects outside the narrow area of you thesis research, then there is no reason why you cannot expand your research interests after graduation.

In conclusion, going to graduate school is a serious decision that needs careful consideration and it should not be taken lightly.

Rodney Brooks interview on the Talking Robots podcast

Marcus Waibel interviewed legendary robotics scientist Rodney Allen Brooks on Friday, April 27th, on the Talking Robots podcast. Brooks is a professor and director of the MIT Computer Science Artificial Intelligence laboratory. He is also the founder of iRobot the company that has introduced a mobile robot in more than 2,500,000 homes; that is the current estimate of Roomba robotic vacuum cleaners that iRobot has sold since 2002. His company is also a major supplier of military robots with their Packbot line of tactical robots.

The Talking Robots interview is very interesting covering some of Brooks’ early adventures in model-based computer vision and his Isaac Newton moment when he conceived the basic ideas that lead to the subsumption robot control architecture. In fact, it was Rodney Brooks in the late eighties who challenged traditional artificial intelligence’s sense-plan-act methodology. The reactive architecture that he proposed did not require the construction of world models which 20 years ago was hard to do because of slow computers and poor mathematical models for expressing the uncertainty in sensing and acting. Today, robotics is driven by AI systems that are a mixture of traditional planning techniques and modern reactive architectures. Bayesian statistical models and Monte Carlo techniques allow intelligent robots to properly model their surrounding environment and plan forward in time to select the best actions for their circumstances. The mixture of the two methodologies allows for reactivity to infrequent but important events but also long term planning something that is not possible using only a subsumption system.

During the interview, Brooks refers to Bayesian statistics (and more specifically their use in AI research) that is popular today as nothing more than another fad that will go away sooner or later. Although, I respect him for what he has achieved in his life, I sincerely believe that this time he is completely wrong. Since his early success with constructing simple purely reactive robots, his group has little to show for. In the meantime, probabilistic robotics techniques are used to operate robots autonomously giving tours in museums, driving across the desert and exploring Mars.

At any rate, Brooks concludes the interview with some predictions about the future of consumer robotics. Interestingly, he makes a good point about the future role of robots in our homes and the cultural differences between North America and Asia. He mentions that in Japan and Korea, they are interested in developing companion robots for the elderly while in North America and Europe, people are interested in developing helper robots that will perform specific jobs for us but won’t be considered our companions.

I strongly encourage anyone to listen to the podcast because Brooks is always a fun person to hear talk about robots. Also, I strongly recommend his last book, Flesh and Machines: How Robots Will Change Us.

Dextre: dual-arm robot completes the Mobile Servicing System (MSS) for the International Space Station (ISS)

The Canadian Space Agency (CSA) announced today that they have completed the 3rd robotic component specifically designed for servicing the International Space Station. The Special Purpose Dexterous Manipulator (SPDM) nicknamed Dextre is a dual-arm remote controlled robot that will be used along with the Space Station Remote Manipulator System (Canadarm2) and Mobile Base System to replace and repair parts of the space station. The robots will perform a vital role in space station operation significantly reducing astronaut Extravehicular Activities (EVA) which is very dangerous due to exposure to radiation and the probability of collision with extraterrestrial objects.

A variety of tasks will be performed by Dextre including installation and removal of small payloads such as batteries, power supplies and computers; providing power and data connectivity to payloads; and manipulating, installing, removing and inspecting scientific payloads. A typical task for Dextre is to replace a depleted (100 kg) battery, which involves bolting and unbolting operations as well as millimetre level positioning accuracy to properly align and insert the spare battery within its worksite and properly engage all connectors. This peg-in-the-hole type of task demands a great amount of precision and a gentle touch to avoid binding. To achieve this Dextre has a unique feature which complements its remarkable dexterity: precise sensing of forces and torques at the "hand" and automatic compensation to ensure the payload moves smoothly into its mounting fixture. To illustrate the level of performance of Dextre, here on Earth it could likely be used to insert an item as delicate as a videotape into a video recorder.

Dextre, much like the Canadarm2, was designed by MD Robotics in Brampton, Ontario. The remotely controlled robot manipulators have been Canada’s main contribution in the international effort of constructing the ISS. According to the CSA press release, Dextre will be transported to the space station in nine pieces in February, 2008, where it will be assembled and go to work shortly after. Remote control of Dextre will be possible by both astronauts on-board the space station and also engineers from Earth.

Image of Dextre is copyright MD Robotics and Canadian Space Agency.

CMU official announces the Telepresence Robot Kit (TeRK)

After one year in use by educational institutions, Carnegie Mellon University’s Robotics Institute officially announced the availability of their educational robotics Telepresence Robot Kit (TeRK.) The idea behind TeRK is to make available simple recipes for building small, Internet controlled robots for fun and education. The team keeps the cost of robot building low by using off-the-shelve hardware components and open source software. TeRK is suitable for teaching robotics in middle school, high school and college.

Unlike other educational robot kits on the market, TeRK is not sold as a complete set of parts. The CREATE Lab’s recipes allow for a variety of robots to be built with parts commonly available through hardware and hobbyist outlets.

At the heart of each TeRK robot is a unique controller called Qwerk that combines a computer with the software and electronics necessary to control the robot’s motors, cameras and other devices. Qwerk, developed by the CREATE Lab and Charmed Labs of Austin, Texas, also connect the robot automatically and wirelessly to the Internet so it can be controlled by any Internet-connected computer.

The fact that the Qwerk is a complete computer running the open source Linux operating system along with the build-in Internet connectivity makes the TeRK standout compared to the very popular Lego Mindstorms NXT kit. The TeRK can easily be expanded with USB devices including Web cameras and GPS receivers; I guess the latter would be useful if you wanted to experiment with an outdoor robot.

Two of the recipes currently available on the project’s website are the Qwerkbot which is a traditional wheeled robot design with differential drive and the TeRK Flower which is a more interesting robot but not mobile. The Qwerkbot will cost you $550 and it requires roughly 2 hours of your time to assemble. The TeRK Flower is a bit more challenging and expensive as it will require 8-10 hours of your time and make your pocket $725 lighter. The cost for both robots is mostly due to the $349 Qwerk embedded computer. The TeRK robots are programmable via the Flower Power and Express-O-Matic visual programming software environments.

Photos are copyright the TeRK project.

Mapping cities using a modified autonomous Smart car

A joint project between the University of Freiburg, Germany, and ETH Zurich, Switzerland, is on a mission to create large scale maps of villages and cities using an autonomous driving car. The problem is an instance of the fundamental Simultaneous Localization and Mapping (SLAM) problem but at a much larger scale in an outdoor environment. The team has outfitted a Smart car with a 5 SICK LMS laser range finder sensors, differential GPS, inertial measurement unit (IMU), and optical gyroscope; data from all these are integrated in a stochastic framework for performing SLAM. The stochastic framework used is based on the theory of Information Filters (IFs) that is closely related to the Kalman Filter (KF) often preferred by the SLAM research community. The information filter outperforms the KF when needing to combine information from multiple sensors such as in this case.

Other than the localization and mapping software, the autonomous vehicle must also drive itself; as a result, it must process the information in the map learnt in order to decide which areas are drivable or not. The maps are augmented with traversability information in order to aid in this task. The resulting map is called a multi-level surface map (MLS.)

The team recently presented astonishing results of their mapping algorithm with the car operating in an unstructured urban environment for long periods of time. Data was collected by driving the vehicle around the EPFL campus for a total distance of 2.3Km. The resulting MLS map constructed with a resolution of 50cm was derived from the measurement of nearly 70,000,000 data points. The area mapped spanned 300 by 200 meters; the final MLS map required 55MB of computer disk storage space. Processing the data was not real-time but only 15 minutes was needed which is rather impressive for the amount of data that had to be processed.

The team is preparing the autonomous Smart car for an entry to the European Land-Robot Trial (ELROB) competition which is the European version of the DARPA Grand Challenge hold the money prize.

Photo of robot car is copyright SmartTer team.

Heavy duty robot: The autonomous hot metal carrier (HMC)

Considering the current trend in consumer-level humanoid robots, one would think that everyone is always trying to build the smallest but smartest robot. There are however applications in which bigger is better and when we say bigger, we really mean humongous!

Researchers at the Autonomous Systems Laboratory, CSIRO ICT Centre, Australia are developing autonomous hot metal carrier robots that are 20-tonne forklift-like vehicles that carry 8 tonnes of molten aluminum in 2 tonne crucibles along a predetermined route from pot lines to casting machines. The team has demonstrated an HMC capable of autonomous operation over considerably long periods of time. The task includes crucible pickup and drop off as well as autonomous navigation in indoor and outdoor environments. Safe operation in such a system is clearly of the at most importance considering that the HMC must share workspace with people but also other expensive equipment. One of the most critical aspects of such a system is reliable localization and obstacle avoidance; the team has decided to use a combination of laser, vision and odometric sensors in a stochastic framework for reliably solving these tasks.

The most recent HMC prototype performed over 100 hours of autonomous operation with little supervision. During one very successful run, the HMC operated continuously for 5 hours traveling 9 km and handling the crucible a total 60 times. Incredibly, a supervisor monitored the system’s operation and provided high-level control as needed remotely via satellite link.

One small step for Nao but one giant leap for French robotics

The French startup Aldebaran Robotics has finally released a video showing their humanoid robot Nao walking. Aldebaran Robotics was established in 2005 with the goal of creating an affordable and intelligent humanoid robot for the public with its principle function being that of entertainment. Project Nao’s goal is to deliver a small humanoid robot fulfilling the above requirements. The prototype robot has seen several iterations starting from AI-01 and ending at the current prototype AI-05. The company that has received funding to the tune of 800,000 Euros has been rather secretive about their robot releasing only scattered information about its design. We know that Nao stands 22 inches tall and has 25 degrees of freedom. In addition, the robot comes equipped with a video camera, speech recognition software and Wi-Fi connectivity. The company claims that the robot will be fully customizable providing lots of fun for the entire family. At any rate, Aldebaran Robotics has finally release a video showing Nao walking.

You can watch the video here.

I have to admit that considering the amount of time and money invested in this project, I am not particularly impressed by the robot’s walking abilities. Similar sized robots available to hobbyists from Korea and Japan are way ahead Nao in their ability to walk smoothly and perform difficult tasks such as skating. Aldebaran claims on their site that Nao will be available for purchase in the second half of 2007 and that the prototype robot shown in the video is not the final product which should be much more improved. I sure hope that they make true to this claim and I am looking forward to the final product which hopefully will be reasonably priced. At least now, Europe can claim that they have also designed a small humanoid robot for the consumer marker; when are we going to see the same from a company in North America or are we so far behind everyone else that we aren’t even trying anymore?

NASA’s tumbleweed-inspired rovers for large scale planetary exploration

NASA has had enormous success in exploring Mars using traditional wheeled rovers. Sojourner was the first rover to successfully land and explore the Martian surface and it was later followed by the larger Spirit and Opportunity rovers that are still operational today. However, the traditional rover design is limited to exploring areas that are nearby the landing site. Even though the larger rovers have explored much more of the planet’s surface than Sojourner, there is still much of the surface remaining unexplored.

In order to explore larger areas that are further away from landing sites, NASA has been working on the concept of tumbleweed rovers that are powered by in-situ resources, specifically the wind. The rover’s design is inspired by nature and as the name implies the Tumbleweed or Russian thistle (Salsola Tragus.) The actual prototypes that NASA has been testing for almost 5 years look more like large “beach balls.” These rovers are lightweight and can carry a number of scientific instruments over long distances. The tumbleweed rover moves by wind power which on the surface on Mars achieves speeds of 2-5 m/s during the day with gusting periods of 10-20 m/s.

NASA has previously used the "beach ball" design during the Pathfinder mission that delivered the Sojourner rover. It was not used in the rover but instead as a low cost landing method. The Pathfinder mission used airbags to cushion the spacecraft’s impact with the Mars’ surface during landing. The spacecraft traveled a considerable distance rolling on the planet’s surface before coming to a stop.

A NASA document published in 2003 proposed a careful study of this new method of rover locomotion including a number of different conceptual designs for a tumbleweed rover. These concepts included a Wedges, Box Kite and Dandelion configuration for the rover. The former was designed to allow a scientific instrument located in the center of the sphere to reach the Martian surface by deflating some of the wedges. The Box Kite concept was focused on the long term survivability of the rover while the Dandelion design was the one most closely mimicking the Russian thistle’s structure and it was optimized for getting the most from wind power.

Over the last few years, NASA has refined and improved the tumbleweed rover concept and they have also performed successful experiments with prototype rovers in the Antarctic. In a recent publication, NASA scientists outlined the feasibility of using a team of tumbleweed rovers for the exploration of the Dao Vallis valley in Mars. Dao Vallis is 1200Km long with gulley features that are believed to have been formed by water making it an interesting area for scientific exploration. To give you an idea as to why better locomotion methods for planetary rovers are necessary, NASA’s twin rovers currently on Mars can only travel a maximum distance of 100 meters or 0.1 Km a day. As you can tell, traditional rovers are not suitable for the exploration of large areas such as Dao Vallis and the Tumbleweed inspired robots are a much better choice.

Images are copyright NASA.

The jumping mini-robot Grillo

During the recently concluded International Conference in Robotics and Automation (ICRA 2007) which took place in Rome, Italy, Umberto Scarfogliero and his colleagues Cesare Stefanini and Paolo Dario presented their prototype long-jumping robot nicknamed Grillo. The robot has four legs and is 50mm large while it weighs a mere 15 grams. This is Grillo’s creators describe the robot’s design and locomotion in their paper,

Inspired by frog locomotion, a tiny motor load the springs connected to the hind limbs. At take-off, an escapement mechanism releases the loaded springs. This provides a peak power output that can exceed several times the maximum motor power. In this way, the actuation and energy systems can be significantly reduced in weight and size. On the other hand, passive dynamics is exploited by compliant forelegs, that let to partially recover the impact energy in their elastic recoil. Equipped with a 0.2W DC motor, the robot is dimensioned to achieve a forward speed of 1.5 m/s, which corresponds to about 30 body length per second.

Jumping locomotion as opposed to legged or wheeled locomotion might be preferable for small robots inhabiting uneven and unstructured terrain. As the researchers explain, at the mini-robot scale, frictional forces dominate and as a result waste a large percentage of mechanical work; this is partly the reason why in nature, smaller animals or insects have developed gaits that require minimal surface contact. This is another case in which nature is a good guide for the design and development of interesting artificial creatures.

Image of Grillo’s 3D model is copyright U. Scarfogliero, C. Stefanini and P. Dario.

Scientists to test a new medical surgical robot controlled over long distances

Researchers from NASA and the BioRobotics Laboratory at the University of Washington have announced their plans to test new technologies in remote controlled surgical robotics. The scientists have constructed a new surgical robot called Raven. The robot is similar to those already found in hospitals but with two very important differences. First, Raven is much smaller than the average medical robot weighing a mere 50 pounds; it was designed to be small and portable for use in the battlefield or in space. Second, Raven is designed to be operated over long distances. Yesterday, we discussed neuroArm that is a new MRI-enabled robot from the University of Calgary. NeuroArm and other similar robots found in hospitals today are designed to be operated by a surgeon in close proximity. Raven is designed to be used over long distances. When there is significant distance between the surgeon and the robot, lag or latency in transmitting the surgeon’s commands and receiving visual and haptic feedback from the robot’s sensor becomes a major factor in the operation’s success. The University of Washington and NASA teams have developed technologies to minimize the delay and also reduce the impact it has to the surgeon’s performance.

As a result, the test that NASA is preparing will involve a doctor operating the robot from coast-to-coast. Raven will be located in the Aquarius Undersea Laboratory off the coast of Florida while the doctor operating it will be located in Seattle. In addition, NASA will be testing the robot’s performance in a simulated zero gravity environment that is underwater. Clearly, NASA has an interest in this technology considering their plans to send humans to the Moon and eventually Mars.

Lastly, this is another project involving the collaboration of scientists and engineers from many different disciplines proving once again that large projects require interdisciplinary expertise in order to be successful.

More on Forbes.com and MSNBC.com

Photo of Raven is copyright the University of Washington

Karto: web-based mapping from laser range data

I received the following email today announcing SRI International's new web-based mapping software for robotics. The mapping software is part of SRI's Karto development platform.

As part of the first step toward providing a complete suite of robotic navigation and exploration algorithms, we are making the Karto Logger Plug-In available as a free download at www.kartorobotics.com. The Karto Logger Plug-In allows robot developers to collect and log odometry and range data in a form that our mapping software can interpret to create maps of indoor environments. This can be done either in simulation or a on a real-world robotic platform. We currently support both the Player and Microsoft Robotic Studio platforms.

We welcome any and all feedback on our algorithms. Please feel free to contact us at kt-info@kartorobotics.com to tell us about your experiences with our mapping algorithm.

Karto works with data collected using a laser scanner that is a popular sensor in modern robotics. Researchers can collect the data by driving their robot around the environment they want to map and then upload the logged data to the Karto Map Server. The server processes the log file and generates a globally consistent 2D occupancy grid map of the environment. An occupancy grid map is much like a floor plan. Once the server is finishes estimating the map, it sends it to the research via email. Check the Karto website for a demo of the mapping software.

Canadian researchers design pioneering robot to assist with brain surgery

Collaboration between the University of Calgary and MD Robotics over the last few years has been very fruitful as the two parties have completed the design and implementation of the neuroArm robot for assisting in brain surgery. The remote controlled device is the world’s first MRI-enabled microsurgical robot. The operating surgeon seats at a remote workstation controlling the neuroArm viewing live 3-D magnetic-resonance images of the patient’s brain.

The neuroArm’s creators claim that this device will revolutionize brain surgery because it will allow doctors to perform procedures with higher accuracy and dexterity. Dr. Garnette Sutherland, a Calgary neurosurgeon and Project neuroArm leader describes why this is so,

The robot sees much more, and that provides the surgeon with greater precision. With robotics, it is possible to see around corners and manipulate in ways that are now not possible.

The robot will assist in a number of different types of surgery including Neuro, Ophthalmological, Plastic and ENT. Surgeons are currently being trained on using the neuroArm in simulation and it is expected that the robot will assist in its first operation later this summer.

Images of neuroArm from the Hotchkiss Brain Institute website.

Big Brother is watching you... with both eyes!

Or he will be soon. With both eyes? Well, yes, because Big Brother needs to see in 3D.

Computer stereo vision has been around for a long time. In the early days it was considered too computationally expensive to be very practical. But thanks to Moore's law it can now be done on any current desktop PC. You can buy stereo cameras for research (Point Grey Research, Videre Design, Focus Robotics). But the question remains - what is stereo vision good for?

Stereo is a heavily researched topic but how is it practically useful? People who read this blog may think the killer app for stereo is mobile robot navigation. Well, yeah, maybe, but thats a few years away. There are other applications such as motion capture or other target tracking systems. But how about high volume applications? Can we expect to see stereo cameras in use around us somewhere next week?

Well - yes! Just head down to the mall. Want to see one sooner? Take the bus!

Currently stereo vision is being deployed in people tracking applications by several companies and is achieving a lot of acceptance in this application. (Here are a few more: [1] [2] [3]) Why? Well, single camera people tracking is usually based on some form of visual motion detection. Visual motion detection typically uses pixel-by-pixel changes in the image from image-snap to image-snap. But what happens when you are tracking people on a sidewalk, and a bus drives past and makes your scene go from bright sunlight to shadow and back to sunlight in a few seconds? Visual change detection fails. Pixel-by-pixel, the whole scene is changing so rapidly that pixel changes due to motion get lost in the chaos.

With stereo vision, however, the two images are snapped at the same time and therefore under the same lighting conditions. This makes 3D extraction from stereo images robust to lighting changes. The result is that you can detect motion with stereo vision by detecting changes in 3D in a lot of scenarios where single-camera methods fail.

While the obvious industry for these applications is security, stereo is also gaining traction in people tracking for marketing analysis - counting how many people enter a store by the hour, for example. Or counting how many people are waiting in line.

It is interesting that so often a technology ends up having the edge to be the killer app in areas where the initial researchers never imagined.

People tracking sequence image copyright Point Grey Research

Christopher M. Bishop's pattern recognition and machine learning textbook is pure gold

Last week, I finally received my copy of C. M. Bishop's new textbook titled Pattern Recognition and Machine Learning. After reading through it for the last few days, I have to admit that this is one of the most comprehensive and up to date textbooks on the subject available today. Bishop who is a research scientist at Microsoft Research Cambridge and holds a Chair in Computer Science at the University of Edinburgh has written a complete treatment of machine learning in 738 pages suitable for young scientists at the advanced undergraduate and early graduate level. Indeed reading this book requires a certain background in math and computer science. However, Bishop starts by explaining the most basic concepts first so anyone with a bit of extra effort could potentially read and understand this excellent textbook; some familiarity with multivariate calculus and linear algebra certainly would help.

As I mentioned earlier, this is a textbook. Each chapter includes a number of exercises including some for which the solutions are available on the book's website; these exercises are clearly marked in the book. Bishop has done an excellent job with this book. He explains many advanced concepts very clearly. The illustrations and examples found in the book further aid in the assimilation of these concepts. The book starts discussing basic concepts such as probability, decision and information theory and then moves on to present regression, classification, neural networks and kernel methods. In the second half of the book, Bishop presents graphical models, inference (exact and approximate,) sampling methods and finally how to model sequential data.

Bishop is currently working on a companion volume which will include example source code implementing many of the algorithms presented in the book along with the appropriate data sets in order to help students better understand all the key concepts. The book which Bishop is co-authoring with Ian T. Nabney should be available in 2008.

IBM creates high performance 3D chip technology for embedded devices and more

I am not an electronics engineer but I get the feeling that IBM announcing today of their new technology for 3D chips is good news for robotics. IBM has developed new technology that allows chips to be stacked one over the other instead of being placed side-by-side on a board. The advantages in terms of efficiency, i.e., lower power consumption and faster communication rates, come from the elimination of long wires connecting the chips. In terms of numbers, IBM claims that data will now have to travel nearly 1,000 times shorter distances using the 3D configuration.

The new technology that allows the stacking of chips is called "through-silicon via" and it is about "etching thousands of holes through thin, silicon layer (wafer) and filling them with metal to create 3-D integrated stacked chips" (from Science Daily.)

IBM will be offering these 3-D chips for commercial use in the second half of 2007 for wireless LAN and cellular communications, i.e., expect smaller, cheaper and more powerful mobile phones. The company has big plans for this technology as it will allow it to place memory chips very close to processing units decreasing the latency to get data to the microprocessor but also increasing the amount of data inflow. IBM plans to outfit their main supercomputer, Blue Gene, with 3-D chips in an effort to increase its processing power and reduce its power consumption.

Don’t be surprised if this technology makes smaller, more efficient, cheaper and more intelligent robots possible in a few years. If you pair the “through-silicon via” technology with Intel’s multi-core CPUs, you might end up having a seriously powerful supercomputer in your living room. The question is, what are we going to do with all that power and is Artificial Intelligence going to advance at an equal pace?

Japan to create guidelines for the safe performance of robots

Following in the footsteps of Europe and South Korea, Japan announced last week that they are assembling a team of experts to create a document titled, "Draft Guidelines to Secure the Safe Performance of Next Generation Robots." Although this can be thought of as a document similar to the Robot Ethics charters that the European and Koreans are working on, in fact, the Japanese are more interested in drafting a very precise and rather bureaucratic document considering that there are lawyers in the panel of experts assembled. The Japanese are interested in a document that uses very precise language to establish guidelines for the safe use of robots. An early draft of the document suggests that all robots report incidents of human injuries (cause by robots) to a central database. In addition, Japan expects all robot manufacturers to conform to the guidelines. I presume that they mean Japanese manufactures but chances are that not robots developed in other countries would be allowed in the country unless the guidelines are met.

You can read a more detailed article about Japan’s robot guidelines draft at the Times Online, including experts from the document.

Cleaning robot specialist iRobot introduces pool-cleaning Vero series of robots

It would appear the iRobot has a mission to keep our living spaces spotless with zero effort on our part. After having incredible success with the Roomba line of vacuuming robots, they introduced the Scoomba floor washing robot and Dirt Dog workshop cleaning robot. Now iRobot is going outdoors and underwater with the introduction of the Verro series of pool-cleaning robots.ow iRobot is going outdoors and underwater with the introduction of the Verro series of pool-cleaning robots.

There are two robots in the Verro series the 300 and 600. The robots differ in more than just the price with the 300 version costing $799 and the 600 version priced at $1, 1999. Yes, the robots are a bit expensive but if you can afford a house with a large pool then you can afford one of these robots. The differences between the Verro 300 and Verro 600 are given in the press release,

The iRobot Verro 300 is designed for use on gunite or concrete and features a powerful hydro-jet system that power washes the pool, deep cleaning pores, cracks and seams. The iRobot Verro 600 is ideal for vinyl, tile and fiberglass pool surfaces and features sturdy PVA brushes that scrub and clean the pool, while the powerful vacuum and self-contained filtration system picks up and traps debris and microscopic bacteria.

The best part about these robots just like all the other robots that iRobot has designed is the ease of use. Basically, iRobot is creating the robot equivalent of the iPod. All iRobot cleaning robots can be operated via one or two buttons making the interface simple enough that anyone can get the maximum performance out of their investment. The robots are also designed to operate in a turn-on and forget mode; you turn the robot on and place it in the area you want to clean and then go away without having to worry about the robot not doing its job or causing damage. This is really the best quality for these robots and in fact it should be the goal for every robot designed for household usage. I believe that iRobot has definitely figured out the recipe for creating useful, robust and easy to use household robots and I have no doubt that the Verro series is going to be as successful as all the others.

Plen will skate into the hearts of 50 lucky people

By now, we are more than used to seeing cute little bipedal toy robots from Japan and Korea being demonstrated in just about any major electronics and consumer show. Akazawa’s Plen is the only one that can stand, walk, pick up small objects and skate! You read that right. This robot stands 22cm tall and weighs just over half a kilogram. It has 18 degrees of freedom and can be programmed and controlled via your computer’s USB port and Bluetooth.

The best news about this is that if you hurry and you can afford its steep price then you can become one of the 50 lucky people who will get the first batch of Plen robots. If you can pay the $2,399 one of these cuties cost then rush over to AudioCubes.com online store and place an order before they are all gone.

In case that you doubt the claims that Plen can skate, here is the video that proves it,

China’s nuclear powered lunar rover prototype

Chinese researchers from Shanghai institute unveiled today a prototype lunar rover that they hope will be selected by China’s space authority for the country’s first moon mission in 2012. The 6-wheeled robot that stands 1.5 meters tall is reminiscent of NASA’s Mars rovers Spirit and Opportunity. The Chinese rover will come equipped with a number of sensors that will allow it to collect and analyze samples of lunar rocks but also construct 3D panoramic images of the Moon’s surface. The robot is designed to move much faster than NASA’s robots as it is capable of traveling 100 meters in one hour. Interestingly, while until now all planetary rovers were designed to get power from the Sun using solar panels, the Chinese are preparing theirs to be powered using a combination of solar and nuclear power. The reason for this comes from the need to keep the robot powered for several days in a row in darkness since lunar nights can last up to 14 days. China and the U.S. are in a race to put a man on the Moon after 2015. NASA of course has already sent men on the Moon with the Apollo missions but going back is considered a stepping stone towards sending astronauts to Mars. Everyone expects that robots will play a very important role before and during these manned missions. The Europeans are also developing several planetary rovers for Mars exploration. The next couple of decades are going to be very interesting for planetary robotics.

Deb Roy’s ambitious Human Speechome Project

An MIT professor is working on a very ambitious 3-year project to understand language development in humans. Deb Roy, Associate Professor of Media Arts and Sciences and director of the Cognitive Machines Group MIT Media Laboratory is recording every moment of the first 3 years in his son’s life. Roy has outfitted his house with 11 cameras and 14 microphones to make sure that he does not miss any of the significant events in the baby’s language development. One year into the project, Roy has already recorded over 120,000 hours of video that is constantly annotated and analyzed by his graduate students; the project apparently collects over 200 Gigabytes of data daily.

The idea behind this project is to collect data about language development in humans and create a model of how early words and grammar are learned. Once such model is constructed, researchers can use it to teach language to robots. According to Roy,

We're primarily capturing the environment of the baby so we can feed that into a learning system on the computer

The project has been running for almost a year and one of the biggest hurdles at the moment is storing and analyzing the vast amount of video and audio data collected. For privacy reasons, there are gaps in the data but this is still the most comprehensive project in childhood development to day. Wired Magazine recently published a very detailed article about the Human Speechome Project.