Introducing Nao

We have mentioned several times (read One small step for Nao but one giant leap for French robotics) the efforts of French startup Aldebaran Robotics to design an affordable and capable humanoid robot. The company has been working on the robot since at least 2005 and the result of their efforts is the Nao humanoid robot which has begun to gain acceptance as a viable research robot with its introduction to RoboCup in 2007 (read New Nao at Robocup 2007.) The company has also developed a simulated version of Nao that can be used with Microsoft's Robotics Studio application.

Aldebaran has big plans for their creation. They want to make it a true companion robot for families. They hope that in the future, Nao will be customizable (they have recruited two Parisian design school designers to give Nao its current form) with the ability to adopt to its environment and learn new skills. I'd say that we are far from having an affordable robot with such capabilities but Nao is certainly on the right path. The current Nao model can walk, get up from ground after falling over, kick a ball, and grasp small objects. It still has a long way to go before it can perform fluid motions like the now defunct SONY QRIO but it should be there in another 3-5 years (this is my widely unsubstantiated guess.)

At any rate, Nao is a wonderful robot and the video below showcases some of its abilities including interaction with people, visual object recognition, object manipulation, and interaction with other robots.

A multi-robot pursuit system for the US army

The US army's desire to develop a multi-robot pursuit system paints a rather bleak future for humanity. After successfully organizing the GRAND and URBAN Challenges with the goal of developing technologies for autonomous driving military vehicles and sponsoring the development of remote-controlled robots for surveillance and also the detection and destruction of Improvised Explosive Devices (IEDs), the army now has published a call for proposals for the development of robotic systems that (as the New Scientist puts it) will be able to hunt down humans like a pack of dogs.

There are many research efforts within robotics in path planning, exploration, and mapping of indoor and outdoor environments. Operator control units are available that allow semi-autonomous map-based control of a team of robots. While the test environments are usually benign, they are slowly becoming longer and more complex. There has also been significant research in the game theory community involving pursuit/evasion scenarios. This topic seeks to merge these research areas and develop a software/hardware suit that would enable a multi-robot team, together with a human operator, to search for and detect a non-cooperative human subject.

The emphasis on the phrase “non-cooperative human subject” is mine and this is the one phrase that really scares me and should scare the rest of you as well. Notice, that the proposal does not specify that the target is a combatant but rather any non-cooperative human. In other words, assuming that such a system is eventually developed (and it certainly will be developed sooner or later,) there is nothing to prevent the army from using it against the average citizen who happens to disagree with the status quo. Experts interviewed by New Scientist agree that it won't be long before these robots are equipped with weapons and their mission upgraded from finding a subject to terminating him/her!

In addition, I find Phase III of the proposal rather interesting and conflicting with the project description quoted above. Specifically, it says,

Robots that can intelligently and autonomously search for objects have potential commercialization within search and rescue, fire fighting, reconnaissance, and automated biological, chemical and radiation sensing with mobile platforms.

People in need of rescue are not normally non-cooperative. If I am inside a burning building with no hope for escaping then I would do everything in my power to help a human or robot team find me!

The research topic of multi-robot pursuit/evasion is definitely interesting but the ultimate application of detecting non-cooperative human subjects scares me more than anything else I have seen recently.

Miniature jumping robot

Efficient locomotion for minaiture robots is a hard task due to the “Size Grain Hypothesis” which simply describes how the rugocity of the environment increases as the size of the agent (robot in this case) decreases. That is the smaller the robot, the bigger the obstacles. To circumvent the inneficiencies of crawling, walking, or running for miniature robots, researchers at EPFL are exploring jumping as a more efficient approach (others have also developed jumping robots.)

The result of their efforts is a miniature jumping robot that weighs 7 grams and is only 5cm tall. Using a clever design, the researchers have created a prototype robot that can jump as high as 1.4 meters which is more than 27 times its own height. The new robot can jump an order of magnitude higher than all other existing jumping robots.

So, how does it work? The short description as given by the researchers in a recently published paper is the following.

It employs elastic elements in a four bar linkage leg system to allow for very powerful jumps and adjustment of the jumping force, take-off angle and force profile during the acceleration phase.

The following video shows the robot in action; the slow motion footage is worth its bandwidth in gold!

Open source Touchless SDK

The keyboard and mouse have become the standard interface devices for computer input and they are not likely to give up the top spot any time soon. Speech recognition has yet to provide a viable alternative regardless of the effort put in it by several large companies including IBM and Microsoft. Touch screens have recently started to gather attention with Microsoft releasing the Surface computer, Sphere computer, and Touch Wall and Apple the much celebrated iPhone. All of these touch interfaces require specialized hardware to work properly and as such are not the most affordable.

Recently, Microsoft developer Mike Wasserman released the Touchless SDK for Microsoft .Net which enables the creation of applications that can receive input via gestures tracked and recognized using input from a cheap web camera. The following video explains the capabilities of the SDK which is available as open source and can be downloaded from CodePlex here.

HAL exoskeleton available for renting in Japan

The HAL exoskeleton which we talked about last year will soon be available for renting in Japan. If you don't remember, Seiji Uchida, a quadriplegic, came 500 yards from the the top of Switzerland's highest mountains with the help of his friend Takeshi Matsumoto who put on Hybrid Assistive Limb (HAL) suit and then carried Seiji on his shoulders to the top of the mountain.

Now, Tsukuba 's Cyberdyne will start mass producing this mini-exoskeleton making it available later this week to those who need it for $2200 a month. I say that this is a mini-exoskeleton because (according to Yahoo! News) it only embraces a person's lower body providing assistance during walking but lucks an upper body component; the complete HAL suit includes and upper and lower body component and if the first version finds success in the marketplace then it won't be long before the complete suit becomes available to those who need it. For the time being, those using the HAL exoskeleton will not be able to lift heavy weights in a similar fashion to Sarcos' prototype demonstrated last year (see a video of Sarcos' exoskeleton in action.)

However, what makes HAL truly amazing is the way that the suit is controlled.

When a person attempts to move, nerve signals are sent from the brain to the muscles via motoneuron, moving the musculoskeletal system as a consequence. At this moment, very weak biosignals can be detected on the surface of the skin. HAL catches these signals through a sensor attached on the skin of the wearer. Based on the signals obtained, the power unit is controlled to move the joint unitedly with the wearer's muscle movement, enabling to support the wearer's daily activities. This is what we call a 'voluntary control system' that provides movement interpreting the wearer's intention from the biosignals in advance of the actual movement.

I am tempted to make an Iron Man joke here but this exoskeleton (and the Sarcos one and all the other ones in development) is going to make life so much better for those with disabilities and the elderly. This is academic research and engineering at their best.

A cheap stereo vision camera

Stereo vision cameras have certainly not gained wide adaptation in robotics research to the extend that time-of-flight sensors such as laser have. Part of the reason is the fact that the slower computers of the past have always made it difficult to deploy real-time machine vision systems for real-time robotics. Another has to do with the high price of quality vision sensors.

Surveyor Corporation, a small start-up, wants to change this by developing and selling an affordable stereo vision camera. The Surveyor SVS camera is priced at $550 which is about 1/3 of the cost of other leading stereo cameras offered by market leader Point Grey Research and Videre Design (even more expensive than the former.)

Surveyor Corporation, a leading developer of small vision-enabled wireless robots and robot controllers for research and education, today introduced a first in the industry by offering a $550 stereo vision system, designated Surveyor SVS, for robotics, embedded image processing and web-based remote monitoring. With onboard programmability, WiFi connectivity, easy interface to sensors and actuators, open source architecture and a low price, the SVS represents a breakthrough in 3D stereo vision technology cost and usability for researchers, educators and developers.

The introduction of this more affordable stereo vision system is definitely good news because it will give hobby roboticists an alternative to IR and Sonar sensing (albeit, the stereo camera is still vastly more expensive than IR and Sonar but the gap is closing.) Moreover, it will make it possible to use stereo vision in an educational setting. On the other hand, I really don't think that anyone doing serious research in robotics and computer vision will select Surveyor's SVS over the Point Grey Research and Videre offerings which may be more expensive but for good reasons. The quality of these cameras and reliability of their software will always be preferred by those with a budget to afford the higher price.