Pelvis pitched forward after switching modes

[Steven Gray]

I transitioned into and out of the joint position control mode and afterward the robot's pelvis was pitched forward about 20 degrees. Is there any way I can reset this? I tried setting the COM height and torso pose; there doesn't seem to be a way to set the pelvis pose.

EDIT: I can repeat this by putting the robot in position control mode, moving the legs around a bit, putting him in walking mode, then setting him down. He will resume balancing, but his pelvis will be tilted.

[DouglasS]

@stgray This seems odd and I have yet to see our robot do this; we use this switch all the time. The best way for us to debug this is for you to submit a log to the WebDAV portal.

[Georg Wiedebach]

On state switch the robot reinitializes the desired joint angles to their current positions. Because of small control errors that causes the robot to drift when switching around between the states a lot without sending new desired joint angles. Upon taking the first step in walking control the pelvis orientation should reset to be level again.

If that is indeed the cause and if you need it we can expose a pelvis orientation packet to the ROS layer. (COM height and chest orientation packets will not influence the pelvis) This packet is a little unpredictable and unsafe though. We did not do that before for that reason. Also, the orientation will be reset every time you take a step.

[Steven Gray]

Okay, that works as is. No need to expose anything else. I admit that I didn't try walking afterwards when I saw it happen. Thanks!

[DouglasS]

@stgray Safe to call this resolved?

[Steven Gray]

Yes, I'd say it's resolved.