Lem Fugitt of Robots Dreams posted this great video of the “Korea Team” putting their humanoid robot through a sword kata at RoboGames 2013. (Note the change of grip about 30 seconds in.)
A British firm called Engineered Arts has been building full-sized humanoid robots that are surprisingly social.
Founded in 2004, Engineered Arts’s flagship product is a full humanoid called RoboThespian. RoboThespian’s head features a moving jaw, internal colored lights, and a small LCD screen for each eye, which allow for substantial expression. The robot has over 30 degrees of freedom, powered by unique hybrid pneumatic-electric actuators. The movements (which you can see in the video below) are surprisingly fluid — I find them reminiscent of the animatronics at Disney, but unlike those, RoboThespian also contains sensors and can be programmed to interact with people in a variety of ways, including speech. Read more…
A new start-up called Entropica is claiming to have discovered mathematical equations that allow an autonomous system to select and achieve its own goals.
Entropica is a powerful new kind of artificial intelligence that can reproduce complex human behaviors, including the ability to autonomously set and implement its own goals. In this video, we will see how Entropica can walk upright, use tools, cooperate, play games, make useful social introductions, globally deploy a fleet, and even earn money trading stocks, all without being told to do so.
Here’s the full pitch video: Read more…
Humanoid robot hobbyist Michael Overstreet — that is to say, Michael Overstreet, whose hobby is humanoid robotics, not that he’s a humanoid himself — though of course he is, as all humans are humanoid by definition and Michael is certainly one, but — let’s start over.
Michael Overstreet has achieved prominence in the U.S. humanoid robotics scene via his Bioloid-based robot “Boomer,” and is now taking it up a notch by trying to 3D print his own DARwIn-OP. The Darwin is a fairly high-end research/education robot based on twenty Robotis MX-28 servos. Those servos are quite pricey, so the robot as a whole is still going to cost him about $6k — but that’s about half what an off-the-shelf Darwin costs. Read more…
One of the humanoid events held at RoboGames this year was autonomous weight-lifting.
This event isn’t entirely new to the RoboGames venue; it and related events were first introduced in 2007. But they haven’t been very consistent about it, either — the couple of times I attended RoboGames in the last few years, we were lucky to get someone to actually run the standard Robo-One style competitions, and never mind anything extra.
But apparently more effort was put into the humanoid events this year, because robotics reporter Lem Fugitt caught some cool videos of the weight-lifting competition. First, here’s an interesting-looking humanoid from Korea called RnD_Eska: Read more…
Long-time futurist and artificial intelligence (AI) researcher Ray Kurzweil (bio, wikipedia) has been appointed Director of Engineering at Google. Kurzweil was reportedly impressed by the remarkable progress Google has made in a branch of AI called deep learning. Deep learning is the new buzzword for hierarchical machine learning techniques, that learn both low-level features (such as edges and corners in a vision application) and higher-level concepts (such as kittens or faces).
Buzzword or not, deep-learning techniques have produced some impressive results in recent years. For example, the latest version of Android OS uses these techniques to dramatically improve the speech recognition on its phones, to a level comparable to Apple’s Siri — except that while Siri sends your voice over the network to be interpreted by big servers in the cloud, Android can do it locally on the device (at least in the case of “voice typing”).
That brings me to the reason why all this is relevant to robotics. While it seems like a lot of the progress in AI lately is being done with supercomputers (like IBM’s Watson) or truly giant networks of servers (Google), that’s only the beginning. Throwing computing power at an AI problem certainly helps, but once the techniques are worked out, most of them can be scaled back down to the level where they will run on smaller, robot-sized computers.
Of course, there are also those who think “cloud robotics” is the future — that is, robots with very little intelligence onboard, which instead rely on server-side processing.
In either case, when Kurzweil’s drive to create human-level AI is combined with Google’s deep pockets, army of computer scientists, and massive computing infrastructure, I predict some exciting advances in the next few years. Stay tuned!
[Via MIT Technology Review]
Providing a good sense of touch has been one of the greatest challenges in robotics — especially in hobby robotics, where budgets are limited and environments are chaotic. There are force-sensitive resistors, but these require fairly large amounts of pressure to produce a reliable signal. Then there are capacitive touch sensors, but these only sense touches from certain kinds of things (like fingers), don’t provide any pressure reading, and generally go haywire if they get even a little wet. Read more…
