PDA

View Full Version : RX-64: Just one more weapon in Skynetís arsenal



DresnerRobotics
04-30-2009, 05:13 PM
I remember being blown away at how strong my first hobby servos were when I built my Hexapod 2 from Lynxmotion about 12 years ago. In comparison, they were literally just a fraction of the strength of modern digital servos, and the gap widens even further when put up against a robotic actuator (http://www.trossenrobotics.com/store/c/2698-Robot-Servos.aspx?a=blog) such as the RX-64 Smart Actuator. (http://www.trossenrobotics.com/dynamixel-rx-64-robot-actuator.aspx?a=blog)

(http://www.trossenrobotics.com/dynamixel-rx-64-robot-actuator.aspx?a=blog)
I was working on my mech Hagetaka (http://blog.trossenrobotics.com/index.php/2009/04/16/hagetaka-a-bipedal-combat-robot/) the other night and made the mistake of grabbing at the robot to stabilize it while it was moving and managed to graze my finger in one of the joints. It drew blood and immediately reminded me that working with these types of servos was an entirely different ballgame than your standard hobby servo. With that in mind, we put together a little demonstration video of just how powerful these servos can be! Enjoy!


YouTube - RX-64: Just one more weapon in Skynet's arsenal

crabfu
04-30-2009, 08:50 PM
Wow, is that the jaws of life for Mech Warefare? impressive....

Adrenalynn
04-30-2009, 10:31 PM
Wow, is that the jaws of life for Mech Warefare? impressive....

Nope.

Jaws...Of... DEATH!
:tongue:

Noodle
05-01-2009, 08:29 AM
Crazy. I happen to like my fingers.

Bullit
05-01-2009, 05:41 PM
Those poor chickens :sad:

DresnerRobotics
05-01-2009, 05:52 PM
I had nothing to do with the video itself! I just filmed the intro and setup the RX, blame the interns! :p

ScuD
05-01-2009, 06:24 PM
If the price on these puppies ever went down, we'd need a new law on illegal weapons... crazy stuff.

Adrenalynn
05-01-2009, 09:43 PM
Yeah, because there should be a law against killing cooked chicken and ground up pig-parts!

Matt
05-01-2009, 11:04 PM
FYI - this video is now #3 on youtube for most viewed in technology for today :) cool.

Adrenalynn
05-02-2009, 12:48 AM
Genius viral work. That's the kinda thing that is has like GeekMusk sprayed all over it. Given my new job title, I'm gonna have to come to you guys for viral marketing ideas. I'm afraid I can't make that stuff up, but darn it if I don't know it when I see it. Grats!

Alex
05-04-2009, 09:07 AM
46,000+ views

woohoo!

DresnerRobotics
05-04-2009, 09:49 AM
Obviously its the big pretty yeti of a man in the video that skyrocketed the views so high :D

dsahm
08-13-2009, 11:22 AM
Question sort of pertaining to this:

On page 29-30 of the RX-64 manual, "...when the maximum torque is generated but the load is too big for RX-64 to move, so that RX-64 ends up in the holding state."

I'm trying to determine what this holding state actually is. My intent is to emulate joint stiffness, where the servo resists someone turning the motor with a torque output proportional to how far they have rotated the motor. SO, if I have a simple 1 fixed mounted servo with a lever arm (initially horizontal), and I want the servo to allow the user to press the lever while resisting the motion with a torque proportional to how far the lever has been pressed.

All thoughts welcome. This would help in understanding how to program fail-safe mechanisms to avoid the finger crunching shown above.

lnxfergy
08-13-2009, 11:37 AM
Question sort of pertaining to this:

On page 29-30 of the RX-64 manual, "...when the maximum torque is generated but the load is too big for RX-64 to move, so that RX-64 ends up in the holding state."

I'm trying to determine what this holding state actually is. My intent is to emulate joint stiffness, where the servo resists someone turning the motor with a torque output proportional to how far they have rotated the motor. SO, if I have a simple 1 fixed mounted servo with a lever arm (initially horizontal), and I want the servo to allow the user to press the lever while resisting the motion with a torque proportional to how far the lever has been pressed.

All thoughts welcome. This would help in understanding how to program fail-safe mechanisms to avoid the finger crunching shown above.

I can't seem to find this in the RX-64 manual (at least the 6-28-06 version on the TR website), got a link to the version you are using? Generally, if the torque overload is on, and you can't move the RX-64 to it's final position, I think it would overload and shut off the torque. I'm wondering if this holding state is some sort of a feature that was removed?

The main problem with doing such a thing as you describe is reading back the torque level. My understanding is that the torque overload is sorta fuzzy, that they are using a current measurement to determine "torque". It also only updates at about 7Hz (at least on the AX-12, but I believe the code in the onboard AVR is probably quite similar).

I think the best fail safe to avoid finger crunching is probably to keep fingers away. Possibly putting a skin over the bot to keep anything out of the joints? I know the farell's put kids clothing on their EX-106 bots, I wonder if that sort of a thing would be enough to keep hands out of joints...

-Fergs

dsahm
08-13-2009, 12:12 PM
I am reading v1.10. Here is a link: http://www.crustcrawler.com/motors/RX64/docs/RX-64_Manual.pdf


The main problem with doing such a thing as you describe is reading back the torque level. My understanding is that the torque overload is sorta fuzzy, that they are using a current measurement to determine "torque".

Does this imply that the Torque Limit ability is also not reliable? I am referring to the address 34,35 locations that seem to limit the output torque of the motor.

lnxfergy
08-13-2009, 12:17 PM
Does this imply that the Torque Limit ability is also not reliable? I am referring to the address 34,35 locations that seem to limit the output torque of the motor.

It will cut off the torque when the load is too high, and the comparisons between torque/current are pretty good I believe, but it may not cut off instantly. This feature is intended to avoid stripping gear trains, or burning up motors, not to implement compliant-manipulators (really, stock servos don't do compliant manipulation).

Edit: after a quick discussion, I realized I need to elaborate: the Dynamixels do have some compliance, and you can adjust how compliant they are, but it is always in the neighborhood of the goal position. For instance, if you set it to 512, but have a high compliance, the servo may end up at 490 and stay put ... the idea here is to reduce the constant jiggling of the servo, and the heat/noise that produces -- not to build fully compliant manipulators that can be pushed any which way.

-Fergs

dsahm
08-13-2009, 01:58 PM
I suppose the next logical question is what type of device is good for compliance control and high torques? I am going to continue pursuing some work-arounds to try to emulate some stiffnesses. I have one idea that would involve exploiting the high refresh rate that the servos have. I could adjust the goal position at such a fast rate that even though the motor is actually switching between free-motion and hold-position, the speed would be fast enough to be perceived as compliance.

roycepipkins
08-22-2009, 12:11 AM
IIRC, DC motor torque is a function of current. I think what you'd want for your stiffness concept is a constant current power supply mode (max current, really) for the servo where reducing average voltage to obtain the average max current, in favor of obtaining the goal position, is not considered an error condition.

This is something I've thought of from time-to-time as well. It'd be neat to have a servo arm work something like helping hands. It could also be handy for posing robots. While the servo is pegged at max current, you'd want the goal position to be automatically adjusted as the servo is forced into a new orientation. I'm not sure you'll be able to get that out of the Dynamixel.

One of these days I want to take a swing at making an openservo (http://www.openservo.com/). As the code is published, such a mode could potentially be added.