Swimming Cockroach Robot Developed

Onnimikki writes "The Ambulatory Robotics Lab at McGill University has made a six-legged swimming cockroach robot as part of Project Aqua. The robot is a waterproof version of the RHex robot, whose inspiration is the biomimetic work by Bob Full of Gecko glue fame. Other cool stuff from the ARL page includes a waddling bipedal RHex, and the world's first galloping robot."

It runs QNX

[leeroybrown]

It's nice to see that it runs a proper Real Time OS.

I have actually seen one case of someone trying to build a mini sub-aqua robot running Windows XP (yes XP not CE) on a powerful micro PC card.

Seriously, ... it sounds fscked up, but it's true.

original rhex is equally impressive

[Anonymous Coward]

Yes, the swimming robot version is cool, but the original rhex robot is pretty incredible too. I work in the lab at UMich where they're working on the land-based one. A friend of mine used a learning algortihm called Amoeba (sort of a hill-climbing approach using simplexes) to speed it up dramatically. It runs fast, much faster then you would expect a stocky little robot with six legs to run. Currently, they're working on a vision system so it can track objects and follow lines, and having it sense its terrain and modify its gait accordingly. Not your daddy's robot!

Already done better

[Anonymous Coward]

In this year's Technogames (in the UK, broadcast on TV by the BCC), there was a robot that could swim underwater. It swam like a fish, with horizontal tail movements, and knew when to move up or down to stay slightly below the surface of the water. It was autonomous, not remote controlled. Much more impressive than the movie of the swimming cucaracha.

Complicated much?

[pitboss8881]

Whatever happened to the wheel? You know that wonderful invention that converts rotational motion into linear motion. Hey, our offroad vehicles use it. Are our robots too good for such antiquated ingenuity? Is the answer simply too easy to give the robotics community the type of intellectual hooplah they thrive on? Or is there some technical reason why trying to make a robot walk is better than letting it roll?

Re:Complicated much?

[nvalid]

There are at least two advantages to legged robots that I'm aware of (though the technology is not necessarily there to take advantage of them). The first is the potential to climb much larger obstacles for a robot of a given size. Robots using rocker-bogie wheel systems such as the Mars Pathfinder vehicle have amazing climbing abilities, but this is nothing compared to what a human can accomplish.

The second is a potential energy savings. Imagine a wheeled vehicle traveling over rough terrain. It's constantly climbing over obstacles which takes energy that is just lost when it falls down the other side. Meanwhile, a legged robot can keep its body above the height of most obstacles and just step over the top of them -- more of its energy goes towards its forward motion instead of the up-and-down motion of the wheeled vehicle.

Oh, another thing is the ability to tolerate loss of an actuator. If one of the wheels were to stop working on a wheeled vehicle, the rest of the wheels would have to drag that one along. Meanwhile, there's been some neat work showing the robustness of legged robots to such problems by groups such as the Biorobotics Lab [cwru.edu] at Case Western.

In the end though, it depends on your application as to which is best. I just can't see that one approach could be better than the other in all cases. Just as one example, I think legged robots have really cool potential for planetary exploration for the reasons given above, but certainly anything spending most of its time on flat ground (agricultural equipment, anything on-road, etc) would perform better with wheels.

Re:Complicated much?

[Dylan Zimmerman]

Robots with wheels certainly can go up and down stairs. They just have to use a strange arrangement of wheels. Specificaly, triangles of them with one wheel at each vertex. When one wheel encounters a pothole (or in this case the gab between two steps' edges), the entire assembly rotates. It works rather well. I've seen a robot designed like that climb about 20 stairs in the same time that it would take to go about 15 feet. The rotation slowed it down a bit, but really, it was pretty efficent.

And the robot from MIT that can do facial expressions is called Kismet. Quite a cool idea, but it can't do much without a mobile base. Right now, it just sits in a room and waits for people to interact with it.

In my opinion, Cog is much more impressive because it learns. They showed it how a toy car works and it learned how to move the car very quickly. Cog actually learned that the car would only move if it is pushed on the front or back, not the sides. Now that's an acomplisment.