Setup
- Robot: several UR5e units, e-Series.
- PolyScope: reproduced on both 5.19 and 5.21.
- We log over RTDE: actual_TCP_force, ft_raw_wrench, actual_current, actual_qd, actual_q. (We use the ROS driver, but the question is about the RTDE/controller fields themselves.)
Observation
When wrist_3 moves, the Tz component (torque about the tool/sensor Z axis = the wrist_3 axis) shows a clear systematic component that does not behave like an external load:
- Tz tracks wrist_3 motor current (actual_current[5]) closely, and follows current more than it follows joint velocity.
- Tz shows essentially no relation to wrist_3 angular acceleration, so it is not tool inertia.
- The same component is present in ft_raw_wrench (payload-uncompensated), so it is not introduced by the payload/gravity compensation in actual_TCP_force.
Taken together, this looks like a coupling proportional to wrist_3 drive current/torque, rather than a real mechanical torque the sensor is measuring. We cannot determine the exact mechanism from the outside.
What we have checked
- Tool / payload: the Tz pattern is essentially unchanged whether or not a gripper is mounted — so the tool’s weight, inertia, and any tool-side mechanical reaction are not the source.
- Static sensor health: with all joints clamped (flange pointing straight down), Tz is sane — no static bias / cross-axis / gravity-model issue.
- Firmware / unit specificity: reproduced on multiple UR5e units and on both PolyScope 5.19 and 5.21 — it looks systematic, not an individual sensor or a single-version regression.
- Software gravity compensation: present in the raw field, as above.
Questions
- Is actual_TCP_force / ft_raw_wrench Tz known to carry a wrist_3 current- or torque-dependent term on the e-Series wrist FT sensor (by design, or as a known electrical coupling)?
- If so, is there a recommended way to obtain a wrist_3-axis torque free of this coupling?
- Is there any documentation of how actual_TCP_force is derived from the raw strain-gauge signals?
Attached is a plot that shows Tz vs wrist_3 current, Tz vs acceleration
Thanks in advance for any info!
