Hi, I need some expert opinions. I bought a few XM430-W350 Dynamixels to make a robot arm. After I powered the robot to make the joints fixed at specified angles, I noticed that when I moved one end of the robot while holding the other end, the joints were not rigid (please refer to the video). Even I moved the links back and forth very lightly and slowly to test each individual joints, the joints still turned slightly. This has happened regardless of whether the robot was powered or not. What do you think is wrong? Is it due to bad backlash?
Yes, that looks like backlash to me.
Can you feel it “hitting” when it’s at either end of that movement?
The solution to backlash is to add some preload. When the robot is standing, the preload will come from gravity. If it’s hanging in the air, you could use a torsion spring.
Thanks. What do you mean by “hitting”? I know backlash is unavoidable but is this kind of backlash reasonable and normal for the higher-end X-series Dynamixel? According to the advertised specs, the backlash is 0.25 degrees. Seems to be bigger in practice.
I have uploaded three more videos on how the motors behave without power:
By “hitting” I mean there’s a fairly obvious difference between being engaged in the gears, and not being engaged during backlash. While backlashing, there should be approximately no resistance; you should be able to basically blow on the parts and they move (if they’re balanced.) Once you hit the end of backlash, and the teeth engage, the resistance should be significantly higher.
0.25 degrees is 15 arc minutes, or about 1/1200 of a rotation. If you measure how far the legs turn during the “shaking” bit, you’ll probably find that the amount of movement divided by radius divides out to about that amount of motion.
Thanks. I see. Here are the close up videos of no resistance (no hitting) for your reference. Is such behavior still typical among the latest Dynamixels and currently available small servo motors from other brands?
https://youtu.be/utQyWFCMuBg (one joint)
https://youtu.be/jwh0EOUVPTI (another joint)
https://youtu.be/vWvoytNA78U (whole robot arm)
Yeah, that looks about right.
You can’t ever get rid of backlash entirely, but more expensive servos, like high-end ones used for industrial control, or the Dynamixel Pro line, can use higher-end gearboxes, like cycloid or strain wave (“Harmonic Drive”) gearing. The cost goes up even more, of course.
One of the hardest things for me to get used to when getting into robotics is that the world of physical/mechanical things, with components not made by the millions for consumer use, has a totally different pricing and performance envelope than the mass consumer market.
Thanks for your info.
This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.