RoboFish are old news (and so are RoboTurtles, for that matter), but the Nonlinear Dynamics and Control Lab at the University of Washington has taken it upon themselves to give their robotic fish some schooling in schooling. Their fish are able to communicate with each other underwater using low power, low frequency sonar. It’s not a very high bandwidth way of exchanging information, topping out at about 80 bytes (32 or so numbers) per second. And even at that, half the information doesn’t make it most of the time. But that’s the great thing about schooling behavior: you don’t really have to know what’s going on.

Researchers found that in schools of real fish, most of them are just following whichever fish around them happen to act decisively. If only a third of a group of fish know where they’re going, it’s effectively the same as if the whole group knows. It’s this sort of really simple yet robust behavior that groups of the UW fishbots are trying to emulate, since it makes reliable communication far less necessary. Their first task (coming up this summer) is going to be to autonomously track a remote control shark, since that’s exactly what real fish like to do. If they can get that figured out, schools of autonomous robotic fish could be used to track real sharks, or maybe slightly less dangerous things, like whales. Or nudibranchs. Or humuhumunukunukuapua’a, because a fish with a name like that just has to be worth following. Probably gets into all kinds of shenanigans.

[ UW ]

If you thought these bots were small, Duke University’s microbots are smaller. Much smaller. Measured in microns (that’s millionths of a meter), they’re 100 times smaller than any similar design at 250 microns long, 60 microns wide, and 10 microns high. The floor that the robots are dancing around in the above video is a mere millimeter across.

The tricky part (one of the tricky parts) about robots this small is getting them to do what you want them to do. These microbots move forward in tiny (10 billionths of a meter) but quick (20,000 per second) little steps in response to electrical impulses applied to the surface they’re located on. To turn, a different signal pulls one end of the bot to the surface, creating a pivot point. By slightly altering the way each bot responds to the “turn” signal, a bunch of them can be controlled at the same time, and fancy math can get the group to respond in specific ways.

Look for these bots microbots to be showing up probably a long time from now in a brain cell near you.

[ Duke University ] VIA [ Engadget ]

If you’ve ever owned a pet that wasn’t a fish, my guess is that you’ll agree that the sense most important to your relationship with your pet wasn’t sight or sound, but touch. It’s a little weird, then, that touch (and reaction to touch) has been largely ignored when it comes to interactive robotics. There are of course a few exceptions like Paro and Pleo, and if you’ve ever fondled a Pleo, it’s immediately apparent that the way the robot reacts to touch is what makes it so satisfying to interact with.

Steve Yohan, a researcher at the University of British Columbia in Vancouver, Canada, has taken this idea to its logical conclusion with the creation of the Haptic Creature, a 100% furry robot rabbit that communicates only through touch. The creature has pressure sensors over its entire body, and responds to touching or stroking by making breathing motions, wiggling its ears, or purring. Yohan says that despite the lack of any other expressive features, users were easily able to determine whether the robot was responding in a negative or positive way.

There’s no way to buy a Haptic Creature, so you’ll have to stick with Paro if you want something fuzzy and cute. But my hope is that research like this will get attention of people like Ugobe, who’ll bring us an ice-age edition of Pleo with a thick woolly coat sometime soon.

VIA [ New Scientist ]

In February, we learned about Dean Kamen’s amazing Luke robotic prosthetic arm, which is way, way, way better than the current standard for prosthetic limbs. It looks like DEKA has put more than a little work into the arm system since then, focusing on what is probably the trickiest (and most important) aspect: how the user interacts with and controls the system.

The arm itself has 18 degrees of freedom, which is nearly as many as a human arm has. That’s a lot of control, which is great, but it becomes that much harder for the arm to be controlled. For most people, sensors on the arm read muscle signals from neurons in the upper body to determine what the user wants the arm to do. DEKA’s system adapts the arm to the user, rather than vice versa, and also allows for the use of macros to make common or repetitive tasks easier to accomplish. DEKA is also working on arms controlled directly with the brain, and part of the above video shows just how effective such an approach can be, as a man uses his mind to lift a cup to his mouth and take a drink, and then sets the cup down again without conscious thought. Simply incredible.

[ DEKA @ D6 ] VIA [ Gizmodo ]

This robot, designed by University of Illinois engineering student Lei Tian, doesn’t seem to have a name. So I’m going to give it one: weedbot. Weedbot is the ultimate in solar powered weed destruction machines. GPS guided and totally autonomous, weedbot uses stereo cameras to locate and identify weeds for extermination based on image characteristics stored in an 80gb onboard database. If a weed is found, weedbot not only chops it down, he also applies herbicide to the spot ensuring that the weed will not rise again. You know, like cutting the head off a zombie. Or something. Weedbot is 2 feet tall, 5 feet long, and can travel about 3 miles an hour on wheels or little tank treads. Since he applies herbicide so precisely, weedbot is much cheaper and more environmentally friendly than typical agricultural weed management techniques that involve spraying an entire field.

Weedbot is currently confined to experimental fields at the University of Illinois, but I’m hoping he breaks out and finds his way over to my back yard.

[ U. Illinois ] VIA [ Robot Living ]

After seven minutes of nail biting trans-atmospheric intensity, the Phoenix Mars Lander touched down in once piece Monday evening and there was much rejoicing. Unlike the Mars rovers, Phoenix is a stationary robotic laboratory, and relies on an extendable arm to take samples of the surface for analysis. Phoenix landed way up in the arctic regions of Mars, and is going to be digging around under the soil looking for water ice and (if we’re lucky) little Martian buggies. The successful landing of Phoenix comes (to be honest) as somewhat of a surprise, since Mars has a nasty habit of eating about 2/3 of the spacecraft that have tried to make it there over the years. See one of the first ever images Martian arctic (in color!), after the jump. (Read more…)

Getting robots to walk on two legs at all is certainly not easy, but getting them to walk like humans is harder. When humans walk, we shift our balance in such a way so that we’re constantly falling forward a little bit. It’s an unstable motion that’s very energy efficient, but hard for robots to duplicate since it requires dynamic regulation of movement. Flame, a 1.3m tall humanoid robot developed by TU Delft Ph.D student Daan Hobbelen, has a balance organ, seven motors, and a bunch of springs that all work together to make it the most advanced human-like walking robot in the world (according to its creator, anyhow). I’m not sure how to judge that claim, but even with all his metal bits exposed, Flame’s gait is noticeably humany. Video of Flame walking, after the jump. (Read more…)

This tiny little 7 gram robot can jump an astonishing 1.4 meters, or 27 times its own height. This is a factor of ten better than any other existing robot. It’s able to do this by mimicking insects like grasshoppers, who store energy in springy legs and then release it all at once. This robot uses a pager motor and a bunch of reduction gears to charge two actual springs, and then releases them to catapult itself upward. A small on-board battery provides enough enough energy for over 300 jumps at intervals of 3 seconds.

The great thing about jumping is that it combines the advantages of being on the ground with one of the most important advantages of being able to fly: obstacle avoidance. Wheeled, tracked, and legged vehicles are all at the mercy of extremely uneven terrain, especially obstacles that are significantly higher than the vehicle, and they’re all vulnerable (to some degree) to soft and shifty or slippery surfaces like sand or ice. As this video shows, the jumping robot has no trouble at all getting around:

Yes, it’s not exactly controllable. And yes, it doesn’t exactly land right-side up. But these are minor quibbles, and they’re being worked on. At the moment, the robot is entirely brainless, but researchers are looking to give it a microprocessor, some solar cells, and a sensor package. Swarms of small and cheap robots incorporating jumping technology would be ideal for exploring other planets, or even perhaps areas with hazardous and unpredictable terrain here on Earth.

Like the floor of my bedroom.

[ EPFL ] VIA [ Live Science ]

Even if it’s not immediately apparent, robotics competitions are about research and design in addition to being about fun and awesomeness. Think about it: the best robots generally win, so a big part of any competition is figuring out how to make the best robot. This is why DARPA made the Urban Challenge a Challenge, and why the European Space Agency turned to smart little soccer robots when they were looking for a way to repair solar cells in space:

Cool, right? Don’t forget, there’s gonna be all kinds of robotics competitions at RoboGames next month.

[ ESA ]

Getting robots to stick to things is a surprisingly robust field of research, and we’ve seen all kinds of wicked cool examples of climbing technology, from bots that suck to bots with claws to bots with super sticky gecko feet. SRI International’s electroadhesive robots use an entirely different technology to climb: static electricity.

Somehow, and I’m not entirely sure how, the SRI bots are able to “clamp” to conducting and non-conducting surfaces by generating the same type of electrostatic charge that will stick a balloon to your head if you rub it on your hair. This effect works on surfaces like wood, concrete, drywall, and even glass, and isn’t effected by dust or dirt. Plus, you can turn it on and off instantly. Clamping pressures range from 0.8 to 2.3 pounds per square inch, and only uses 20 microwatts/newton weight held of electricity. Look for this tech to appear in (surprise!) military recon bots, and (actual surprise) maybe even toys or, eventually, human wall climbing accessories.

Click through to the Popular Mechanics article to check out a video of one of these bots climbing drywall, which is about as exciting as it sounds.

[ SRI ] VIA [ Popular Mechanics ]