This video neatly demonstrates the utility of a jumping robot. EPFL’s jumper is simple, small, and cheap, but it’s able to rapidly negotiate an obstacle course that would be otherwise impassible by anything except a flying robot.

The robot plus its self-righting roll cage weighs 14 grams and measures 18 centimeters in diameter. It can jump over 60 centimeters high, which at over four times its own height, is definitely respectable. To steer, the jumping part of the robot is actually able to rotate around inside its roll cage to launch in any direction. Simple but effective.

I remember back in early 2008 when we first posted about this robot, and I wrote:

“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.”

Quibbles solved. Nice job, EPFL. Now just make it fly…

[ EPFL ]

Remember how Chief Cook tried to cheat at pong at ICRA in 2008 by going “hey, look over there, a dancing robot!” Well, now we know what he was pointing at.

If you want more (you want more, right?) head over to Eric Sauser’s website to watch a video of Chief Cook dancing all around Switzerland (complete with table dances) to some sweet, sweet Buffalo Springfield.

[ LASA @ EPFL ]
[ BeatBots ]

In March of last year, we posted about a project to mount WiFi and cellular routers on quadrotor UAVs to enable rapid deployment of networks in disaster areas. EPFL has been working on the same sort of thing, except utilizing swarms of micro air vehicles (MAVs) relying on intelligence algorithms derived from the behavior of army ants. Basically, it’s like LANDroids, except airborne.

SMAVNET (Swarming Micro Air Vehicle Network) consists of a whole bunch of small, cheap (or relatively cheap, more on that in a follow-up post) micro air vehicles. Each one carries an off-the-shelf USB wireless dongle, and by following simple rules to optimize their positions and scout new territory, the drones can spread out, locate a target, establish a robust aerial data network, and then make their way back to base for an automated landing when they’re finished. Sweet concept, right? Well, here’s the system in action using real MAVs:

As far as I know, these 10 MAVs constitute the largest outdoor aerial robotic swarm ever deployed. SMAVNET is primarily designed to facilitate communications in disaster areas. However, it’s hard to ignore the potential military applications, especially considering how similar SMAVNET is to the LANDroid project, which is sponsored by DARPA. There are obviously substantial upsides and downsides to flying network nodes versus ground network nodes, but it strikes me that a combination of the two would be ideal: SMAVNET provides fast response time, while LANDroids offer longer term endurance. That’s just a fantasy, of course, but it’s pretty cool to watch how swarm robotics has been evolving over the last few years, especially now that we’re starting to see real life practical applications for the technology.

[ SMAVNET ]

Thanks Sabine!

We first wrote about Handbot a year ago, when we saw it climb up a bookshelf and steal a book. Handbot relies on a spring-launched magnetic grappling hook to lift itself up off the floor to get its grippers on things up the Z axis… I especially like the cute little propellers which I’m pretty sure the bot uses to maintain its rotational orientation as it rises. Of course, without a detachable grappling hook, Handbot is good for use exactly once, which (obviously) isn’t ideal. Now, however, it’s got a switchoffable magnet, meaning that it can steal two different books. Run for the hills!

The reason that it’s called Handbot is that it’s designed to be the manipulation portion of a robot made up of individual specialized sub-robots, including Eyebots for sensing and Footbots for ground movement. The whole shebang forms “an heterogeneous robotic system” called Swarmanoid, and eventually, one Swarmanoid assemblage will be comprised of some 60 (!) individual Handbots, Eyebots, and Footbots, capable of cooperatively moving around, sensing, and manipulating in 3D space.

As long as the ceiling is magnet friendly, anyway.

[ Swarmanoid ]

I keep on wondering why robotics researchers persist in designing humanoid robots specifically for domestic applications… Quite often, it seems to because they figure if the robot looks like a person, then it’ll be easier for people to relate to it and become comfortable having it in their home.

Such figuring isn’t quite right, and in fact may be entirely wrong, at least according to this 2008 study from the Swiss Federal Institute of Technology and EPFL. Researchers surveyed 240 people at a home and living exhibition in Geneva about their feelings on robots in their lives, and came up with some interesting data, including the above graph which shows pretty explicitly that having domestic robots that look like humans (or even “creatures”) is not a good idea, and is liable to make people uncomfortable.

The location for the survey was chosen because the people attending the exhibition weren’t interested in robots specifically, but rather home technology in general, making them potential early adopters for robots in the home. And since they decided that going to a home and living exhibition was a fun way to spend their time, it’s probably safe to assume that they’d spent some time thinking about what they would and wouldn’t like to get out of a robot. After the jump, more data on what respondents see robots doing for them in the near future.

(Read more…)

Batteries are terribly inconvenient. The more power or endurance you need, the bulkier and heavier the battery has to be, and the more time it takes to recharge. Really, it’s the recharging that’s the problem, since until we develop a feasible ultracapacitor, any battery powered robot is going to have to spend a significant amount of time doing nothing but sitting around recharging its batteries.

One way to get around this is to charge backup batteries external to the robot itself, but that process has generally been more trouble than it’s worth, since batteries tend to be heavily integrated into the structures of robots. Way back in September of 2009, we posted about a conceptual pet care robot that used an external battery swapping method, which was very cool, but it didn’t look like it had a prayer of ever being realized. The video above shows an actual external battery swapping system in action, on a marXbot, which is part of the Swarmanoid project from EPFL. Using a rotary loader, marXbot can swap out its battery in seconds while a capacitor keeps the robot powered. The batteries charge on the loader, so by the time the spent battery makes it all the way around, it’s been recharged and is ready for another robot in need of a fresh meal.

Somewhat ironically, swarms of robots are arguably least dependent on power system restraints, the idea being that you can just have other robots in the swarm cycle in and out to charge. However, the more robots you have, the more charging infrastructure you need. With this battery swapping system, the number of robots that can recharge at once is limited only by the number of batteries in the system, as opposed to the number of charging stations or outlets or something, which is much more efficient.

[ marXbot ]

Robots Podcast episode #50 is not only their fiftieth episode (pretty impressive, right?), but they’ve managed to make it a two parter packed with interviews of twelve different roboticists. Here’s the lineup:

- Rolf Pfeifer: Embodied AI and Robotics
- Mark Tilden: Robot Toys
- Hiroshi Ishiguro: Androids
- Oscar Schofield: Underwater Robots
- Steve Potter: Brain Machine Interfaces
- Chris Rogers: Education Robots
- Jean-Christophe Zufferey: Flying Robots
- Dan Kara: The Robot Market
- Kristinn R. Thórisson: AI
- Terry Fong: Space Robots
- Richard Jones: Nano Robots

You can download the podcast as an mp3 here, or subscribe via iTunes at this link.

[ Robots Podcast ]

Thanks Sabine!

Last time we heard about robot evolution, the bots were figuring out how to deceive each other. Now, researchers at EPFL in Switzerland have been using the same sort of genetic programming techniques to enable robots to teach themselves how to solve mazes, cooperate on tasks, and hunt each other (we’ll save that one for last).

The way genetic programming works is that the robots are only programmed on a very basic level, with simple information on their sensors and objectives. At first, the robots are clueless as to how to take the information from their sensors and apply it to completing their objectives, but after each test, random variations (mutations) are introduced into the code. Robots that demonstrate the most improvement have their code passed on to the next generation, and the process was repeated a bunch of times. In this experiment, after 100 generations, the robots taught themselves how to navigate a maze without running into a wall, and figured out that having their sensors pointed in the direction that they were going was the best way to be. It gets cooler:

“In another experiment they programmed groups of robots to push tokens along a wall to a marked area to win points. They selected the robots that gained the most points to pass their code on to the next generation. Over time altruistic behaviors were observed, in which robots sacrificed points if the entire group would benefit, and the robots cooperated to push larger tokens together to earn more points. As in nature, the robots followed the biological principle of kin selection, in which they only helped robots having the same code lineage.”

Code-based kin selection. Crazy, huh? It makes sense, though, if you think about it… Robots work together best if they’re using the same code, and robots who aren’t using that code won’t know how to cooperate with the robots that are. So, they’ll end up being less efficient, and won’t be as likely to make it into the next generation.

The predator-prey dynamic is perhaps the most interesting. One group of robots with several sensors were programmed to chase after another group of robots, who had fewer sensors but were faster. At first, the predator robots simply chased after the prey robots, who ran away. After 125 generations of evolution, the predator robots had figured out how to stalk the prey robots, hiding out and then sneaking up on the prey’s blind spots. The prey robots, on the other hand, learned where the predators liked to hide and made sure to keep their sensors facing them.

All this, in just a hundred or so generations in a lab. Kinda makes you wonder why we don’t just set a couple hundred Roombas loose in a dust covered wearhouse, let them fight it out and breed with each other, and after a couple years we’d get the world’s smartest, most efficient, and deadliest robot vacuum.

[ Evolution of Adaptive Behaviour in Robots by Means of Darwinian Selection ] VIA [ Physorg ]

The paper also refers to another robot evolution project called Golem@Home, which used distributed computing to design a moving robot from scratch. Different simulations were raced against each other, and the winners were actually created. We posted about it back in 2007, it’s pretty cool.

Remember that jumping grasshopper robot from May of last year? It still hasn’t quite figured out how to fly, but it can now make more than one autonomous jump in a row, thanks to a primitive simple but effective self-righting system. It’s the same type of thing we saw on the WeebleCopter: a spherical metal framework with the robot on the bottom, where its weight will cause the whole thing to roll upright:

There are some downsides to this system, including increased bulk and most notably a decrease in jumping height of nearly 25%, but the frame does protect the robot, and if it gets stuck in a tight spot, it can use the frame to bounce off obstacles and get itself pointed in a different direction.

[ Self Deploying Microglider @ EPFL LIS ]

I can’t find much in the way of additional information on this, but it’s a prototype of a micro air vehicle called AirBurr, and it comes from the Laboratory of Intelligent Systems at EPFL. The version in this video is a remote controlled prototype, but the design is intended to be autonomous, with robust collision recovery in indoor environments. AirBurr is a sort of cross between a helicopter and an airplane, with a bunch of airplane style control surfaces and a horizontal plane lifting rotor like a helicopter.

As far as I can tell, the big advantage of this structure is that it allows the craft to recover from a crash by retaining a wing… Unlike a helicopter, the AirBurr looks like it can take off from the ground in just about any orientation (think post-crash) by using its integrated wing surface to generate lift. I’m just speculating, of course, but hopefully we’ll get some more information in the near future.

What AirBurr reminds me of are R/C airplanes, which have a high enough power to weight ratio that they have no trouble hovering. The downdraft from the propeller moving over the ailerons provides enough force to counteract the torque:

Pretty cool, but I’ve never seen a full size airplane do that.

[ LIS @ EPFL ]