[guidance wanted] Tuning EKI interface for KR 10 R1100-2

[Filip Sund]

Hi,

We have access to a KR 10 R1100-2 robot, and we are currently limited to using the EKI interface for controlling the robot.

We are having some issues with movement execution; specifically there is a big difference between desired and actual joint position. The result being that the robot often overshoots the target joint state, after which the robot "reverses" for a bit, before settling in the target state. The end result is a very uncontrolled movement, and the robot sometimes keeps moving for upwards of 10 seconds after moveit reports Completed trajectory execution with status SUCCEEDED.

This gets especially bad when we try to plan and execute multiple movements in succession. Since moveit reports that the trajectory is complete long before the robot stops moving, the next execution command sends the robot way off the expected path.

We have tried replicating the plots generated by @BrettHemes, but we get nowhere the same performance as he found for the KR6 R700 sixx.

Below are some plots of desired vs actual position when moving joint 2 from -20 to -160 degrees, with a max velocity of 5.24 and has_max_acceleration: false.

Velocity scaling 1.0

Velocity scaling 0.1

Velocity scaling 0.05

Velocity scaling 0.04

Velocity scaling 0.03

Velocity scaling 0.02

Velocity scaling 0.01

It seems like the maximum velocity we can expect for joint 2 (with the EKI interface) is around 5.24*0.05, as it's only when we get to a scaling below 0.05 that the actual velocity starts going down.

Is this as expected, or is there anything we can do to safely have the robot move faster using the EKI interface?

We have tried fiddling with the KRL code and the controller code, without much luck so far.

Any help would really be appreciated.