Ax12/18 pid

[NicoX]

Hi All,
I went through the documentation of AX 12 and 18 and found they have some compliance slope/margin & punch to kinda simulate PID closed loop control. My first question is, does this control algorithm can really emulate a PID control like MX servos give. I guess no... But found some users in forum got a very good results.

So my second q is can any one point me to a tested link where people changed the firmware of AX 12/18 servos to get actual PID control. Google gave me some links but I am not very proficient in doing firmware changing on AX servos. So will it be safe to update firmware or try by compliance process as documentation gives?

Doc: http://support.robotis.com/en/produc...x_actuator.htm
AX 12 PID python code: https://gist.github.com/iandanforth/7185092
Updated firmware: https://actuated.wordpress.com/ax12firmware/

Regards
NicoX

[jwatte]

It is closed loop, but it is not PID. It's more like a "P with some tweaks" controller.
That being said, it's doing just fine for many applications.

Changing the compliance margin and punch of the stock servo firmware is totally safe.
Changing to custom firmware is ... less safe. I'd recommend against it, unless you find that you really can't get the performance you need out of the existing servo with stock firmware, and then only carefully, on one servo, as an experiment.

[NicoX]

It is closed loop, but it is not PID. It's more like a "P with some tweaks" controller.
That being said, it's doing just fine for many applications. Changing the compliance margin and punch of the stock servo firmware is totally safe.

Like it So I can play some to see the end result

unless you find that you really can't get the performance you need out of the existing servo with stock firmware, and then only carefully, on one servo, as an experiment.

Will it be good for a humanoid (20 AX18 servos) controlled by Ordoid XU4 and USB2AX... planning to have Arbotix Pro ?

NicoX

[jwatte]

There are many good humanoids built with stock firmware and AX12 and AX18 robots. (Check out Bioloid GP!) You should worry more about the weight the servos have to carry!

Also, if you use USB2AX, you don't need Arbotix Pro, and vice versa. You should also be careful about the weight of the Odroid and the supporting circuitry (voltage converters?) it may need.

[Le0nX]

Is it possible to achieve smth like this with ax-12a servos? I mean such kind of dynamics.

[jwatte]

Something "like" it? Yes. But not quite that big, and not quite that nice. They created the MX-28 servo specifically to support that robot.

Here's what you can do with AX-18: https://www.youtube.com/watch?v=Ags6Sg05iSY

[tician]

Also this with AX-12/18 servos: https://www.youtube.com/watch?v=DVq8Vp798gg
It is mostly the DARwIn-OP framework and gait engine, but with cheaper servos (https://learn.trossenrobotics.com/38...-overview.html). Unfortunately, it has been discontinued along with the arbotix-pro, but all the aluminum frames are still available at the Trossen shop. The newer CM-whatever controller for the newer DARwIn and Turtlebot 3 should be mostly compatible with the existing software, and I have an arbotix-pro for sale that I pulled from Ripley because I never intended to use the DARwIn-OP framework or gait engine. Also the battery harness and two head turrets/mounts and miscellaneous chassis plates.

The AX-18 has a higher current, coreless motor than the AX-12, so it has a tendency to overheat and shutdown torque under sustained load (trying to hold a fixed position instead of constantly moving back-and-forth). Once a dynamixel servo enters thermal shutdown, it often has to have power completely cutoff before it will actually cool down enough to leave that fault mode.



Hopefully I will have finished revamping all the decades old control electronics at work sometime before next year so that I can start playing with Ripley again and maybe finally take part in MechWarfare. ...unless the economy crashes again... ...or the anxiety/depression kicks my ass again...

[Le0nX]

Thanks for reply! "Not quite that nice" because of lower torque of ax-12a or... because of internal control algorithm of a servo(mx-28 is PID)? What does make the main difference between these two models of servo?

[KurtEck]

I am not sure are you asking the differences between AX12a and AX18 or to MX-28????

Some of these time answers can be found in the emanuals...

http://emanual.robotis.com/docs/en/dxl/ax/ax-12a/
http://emanual.robotis.com/docs/en/dxl/ax/ax-18a/
http://emanual.robotis.com/docs/en/dxl/mx/mx-28/

and the product pages... http://www.robotis.us/ax-series/

Differences in AX12-18 is the 18 can produce a little more torque 1.5->1.8 and more speed 59->97 and eat more amps...
At about double the price $45->$95.

And MX-28T... Other page on products lots different from AX: http://www.robotis.us/mx-series/
$220 - More torque 2.2 more resolution, better processor....

[jwatte]

Yes, the differences between AX-18 and MX-28 are:
- higher-resolution encoder in MX-28
- stronger motor in MX-28
- higher operating voltage in MX-28
- faster embedded control CPU inside MX-28
- MX-28 has 360 degree encoder; AX-18 has 300 degree encoder

If you want to build a robot like the Darwin-OP, you should also look into the XM-W210 and XM-W350 series. http://www.robotis.us/dynamixel-xm430-w210-t/ (this is the "replacement" for MX-28)
(There's also XM430-W350 for slower turn rate / higher toque in this series.)

Also look at the XL430-W250: http://www.robotis.us/dynamixel-xl430-w250-t/
(This is the "replacement" for AX-12)
It has better sensors/control than the AX-12, but it is not noticeably stronger, and thus can't run a bigger robot than the AX-12, just control it better.