Wii Hacked To Control Sword-Wielding Robot

ianchaos writes "WiiBot is the pet project of two engineers who apparently have way too much cool hardware and time on their hands. These two guys figure that as long as you have a Kuka KR16 industrial robot to work with, why not see if you can control it with the Wii Remote? The result is a tennis-playing, sword-wielding mechanical arm that simultaneously captures 'weekend of nerdy fun' and 'accident waiting to happen' in a fun two minute video. The website even details the technical aspects of teaching a robot to parry."

Mirror

[Fred the computer]

Youtube video mirror here [youtube.com]

Re:Military?

[Anonymous Coward]

Uh this is old news buddy. Been there, done that, deployed to Iraq and Afghanistan (at least the robots with guns part). The cool/new thing about this is using a wiimote.

Re:Neat Implications

[Anonymous Coward]

I may be wrong here, but if I'm reading them correctly, they STILL had to manually program the movement sequences (using the robot's built-in 6D controller). Then they had to hand-code roughly analogous coordinates into the software. Finally, they calculated what the user was doing based on the accelerometers in the Wii-Mote, and did a fuzzy match on the 6 actual, 5 practical pre-coded motions, selecting the closest match and performing it.

This meant that the robot could do a handful of simple, pre-defined motions, and the Wii-Mote was simply used to select the closest available match. Not saying it isn't cool, but it's a far cry from programming the robot with a Wii-Mote. I'm not entirely sure those robots could even handle the amount of data it would take to real-time mirror a Wii-Mote. These machines are designed to do a handful of carefully pre-recorded motions (typically one), over, and over, and over, and over for years with near perfect accuracy. Not to mention the fact that there's really no direct way to translate the accelerometer data from a Wii-Mote into useful, sensical motions for a 3-jointed mechanical arm (or any robot, for that matter.) So even if the poor thing could somehow handle that much incoming data, figuring out what data to send it in the first place would be damn near impossible.

What they're doing is cool as hell, but they're not programming the robot with the Wii-Mote. They're controlling it, just like the headlines says. Just sayin'.

Are they really controlling it with the Wiimote?

[Anonymous Coward]

Pay attention when there's 41 seconds to go. The guy is holding his arm up and the robot beats him to bringing it back down. The robot's arm is practically all the way down before he springs into action and brings his arm down. Looks fake as hell to me.

Re:Are they really controlling it with the Wiimote

[Anonymous Coward]

Actually, if you REALLY paid attention, you'd notice that regardless of whatever movements are being made, the robot arm always returns to a default position once the wiimote is no longer registering motion. After every single demonstrated swoop, the robot arm returns to the *exact* same position before they next demonstrated swoop.

Re:Time to Update the Clue Boardgame

[Kierthos]

Um, yes. It's actually played in one of the BOfH stories from.... 2005, I think. Maybe 2006.

Doesn't even need the console

[LiquidCoooled]

The wiimote is a bluetooth device and there are lots of people working on driver software for various operating systems.

You can start at wiili [wiili.org].

Investment cost is about £40 for the mote plus about £10 for the bluetooth dongle.

Re:Are they really controlling it with the Wiimote

[grumbel]

It doesn't look fake, but it does look like they are simply using pre-recorded motions, just like almost all Wii games out there. So instead of the robot mirroring your motion, you have to to mirror a predefined motion to trigger a prerecorded motion in the robot. Looks cool, but it is pretty much useless for actually controlling the robot, since its really no different then pressing the "cool sword swing motion"-button.

The Wiimote can't give you accurate position data, so thats pretty much all you ever get.

Re:Looks like...

[wintersynth]

We actually have a second server at http://www.usmgarage.com/ [usmgarage.com] that will take you to a mirror of the page. If you're going to have your server beaten to death with HTTP requests, the Slashdot crowd is not such a bad way to go.

Re:RUR-tastic...

[TubeSteak]

Loved the discussion of how the bot can easily decapitate.

I didn't notice till the end, but this is the same Aaron who did The Quintessential Sentry Gun [slashdot.org]

Here's an article about him from 2005 [bizjournals.com] which talks about the new company created in order to help him develop the sentry gun. Because the President of that new company is also the owner of a separate automation and robotics company, Aaron gets to play with lots of expensive gadgetry.

Re:Neat Implications

[RimfireShooter]

Real time mirror maybe not but you could get it to duplicate a motion w/o a pre-programmed path. A robot move usually consists of a X,Y,Z location within the operating envelope, a max speed, and whether the move is joint or linear (linear moves coordinates all axis to move along a straight line, a joint move allows the controller to select the optimum path & axis to reach the point)

You could in fact translate the Wii output to useful robot coordinates and velocities that are within the normal robot operating envelope & joint speeds. Simulations are routinely used to generate robot programs to reduce teaching time and a Wii translator program would essentially do the same thing. How to get those Wii translated programs into the robot controller and running them would have to be resolved. You may be able to write your entire program using indirect point references and then modify those dynamically.

BTW - that robot and all recent industrial robots that I am familiar with have 6+ axis of freedom and not 3. Count the servos - there are six.

Re:Are they really controlling it with the Wiimote

[grumbel]

The Wiimote gives you two kinds of data, the coordinates of the IR LEDs of the sensorbar in 2D, which used for pointing, the x,y,z accelerations. It actually doesn't give you exact orientation, you have to derive that from the accelerations, which only works as long as you don't move it, else acceleration and gravity will overlap and you will have a hard time telling which is gravity and which is movement of the Wiimote. Also the orientation you get that way is limited to X and Y axis only, the Wiimote can't detect rotation around the Z axis via the accelerometers, however to a limit extent it might be possible to get it from the sensorbar.

To make it short: I believe it when I see it. So far most Wii games used prerecorded motion, aka glorified button presses. Some games, such as Wii Sports, also take the speed into account, but those only work because the motion itself is very limited. Real 1:1 mapping just doesn't work with the sensor in the Wiimote, you can however of course get a lot closer to it then Zelda, which really was just lame in terms of input.