I have found that the values of the TCP force, taken from the RTDE actual_TCP_force on a UR5 with CB3, are not consistent when the robot is in Teach mode. More precisely, we mounted an additional force sensor (ATI Mini45) on the end-effector and a handle after the ATI sensor, then compared the values taken from the ATI sensor with those received via RTDE. The two values are quite similar (those from ATI are a less noisy) when the robot is not in Teach mode, but if Teach mode is enabled and the robot is moved grabbing only the handle after the ATI sensor, the actual_TCP_force from RTDE remains close to zero, is much more noisy than in the other modes and, of course, is completely different from the ATI measurement.
It seems that in Teach mode the TCP force estimation is somehow “disabled” (even though not totally zeroed…), is someone aware of this and/or can tell me why?
In freedrive, since the robot is “free driving”, the robot tries to create a 0-force vector.
Hence the robot will appear somewhat “weightless” .
So that is probably why you are reading close to zero.