Robot Gets Stuck When Fully Extended

[AlexanderKhazatsky]

When you extend the robot out all the way with the impedance controller as in the picture below, it is frequently (but not always) unable to pull itself back up. It stays in the position with the first 4 joints unable to pull it back up.

[exhaustin]

Our impedance controller is not expected to work well at the edge of the reachable space, as the poses are generally singular.

For more technical details, our controller design is mostly an operational space controller to achieve similar performance in different poses, with a joint space controller component with relatively low gains to deal with joint friction. The problem with the operational space controller is that it will be very hard to move the arm in certain directions under singular poses such as the one depicted in the photo.

If possible, I would try to constrain the experiment to be within a more conservative region. If that's too hard to do, one solution will be to increase the joint gains here. The default values can be found here. Depending on how much you increase them, the arm might be noticeably stiffer (lower compliance) at certain poses (especially at poses close to the base).