iiwa External Force Measurement - High Inaccuracy

mike2902 · August 8, 2018 at 1:26 PM

Hello,

I have started working with a LBR iiwa. While trying to use movement stops with ForceConditions the robot did not start the movement on low forces (e.g. 0.5 N). After looking at the measured force values (see code below) I get calues of up to 3N and inaccuracy of almost 4N (see image). I had nothing attatched to the robot and did nat attatch anything in the programm.

Are these values normal? If not, any ideas on the reason?

Code

ForceSensorData KraftData;


		double Kraft_X;
		double Kraft_Y;
		double Kraft_Z;
		double KraftFehler_X;
		double KraftFehler_Y;
		double KraftFehler_Z;
		double Moment_X;
		double Moment_Y;
		double Moment_Z;
		double MomentFehler_X;
		double MomentFehler_Y;
		double MomentFehler_Z;


		robot.move(ptp(Math.toRadians(0), Math.toRadians(35), 0, Math.toRadians(-85), 0, Math.toRadians(59), Math.toRadians(-45)).setJointVelocityRel(0.5));


		KraftData = robot.getExternalForceTorque(robot.getFlange());


		Kraft_X = KraftData.getForce().getX();
		Kraft_Y = KraftData.getForce().getY();
		Kraft_Z = KraftData.getForce().getZ();


		KraftFehler_X = KraftData.getForceInaccuracy().getX();
		KraftFehler_Y = KraftData.getForceInaccuracy().getY();
		KraftFehler_Z = KraftData.getForceInaccuracy().getZ();


		Moment_X = KraftData.getTorque().getX();
		Moment_Y = KraftData.getTorque().getY();
		Moment_Z = KraftData.getTorque().getZ();


		MomentFehler_X = KraftData.getTorqueInaccuracy().getX();
		MomentFehler_Y = KraftData.getTorqueInaccuracy().getY();
		MomentFehler_Z = KraftData.getTorqueInaccuracy().getZ();

Display More

razzo · August 8, 2018 at 2:37 PM

The force is not measured directly like it would be if you have mounted a force/torque-sensor at the flange. The accuracy highly demands on the robots position. The closer the robot is to a singularity the higher is the inaccuracy. But I think that this is already mentioned in the manual.
But 4N is not very much this are about 400g right at the TCP. You can try to press a dired force on a measurement device and you will see that this is quite accurate. Even other robots with a mounted foce/torque-sensor have almost the same accuracy. The one I measured has even a way higher overshoot.

mike2902 · August 8, 2018 at 2:59 PM

Thanks for the reply. We need to measure contact forces of less than 1N. I know, this is a very low force. An external senor of higher accuracy will be needed

rreype314 · August 9, 2018 at 5:26 AM

The torque sensors are very sensitive. I have also found nonzero values in the external torque and force measurements. My solution so far is to poll the sensors before the start of the motion that you are actually interested in recording, and subtract it.

mike2902 · August 9, 2018 at 11:42 AM

Quote from rreype314

The torque sensors are very sensitive. I have also found nonzero values in the external torque and force measurements. My solution so far is to poll the sensors before the start of the motion that you are actually interested in recording, and subtract it.

So you use a measurement beforehand and use it as an offset that is substracted. Does the offset stay constant during the motion? / Does your corrected value stay at 0?
I did not check this idea yet.

DrG · September 5, 2018 at 3:13 PM

Hi,

Quote from rreype314

The torque sensors are very sensitive. I have also found nonzero values in the external torque and force measurements. My solution so far is to poll the sensors before the start of the motion that you are actually interested in recording, and subtract it.

this is a valid idea, since most of the times, you are interested in the Change of contact Forces...
... there you find a good Resolution in the iiwa, but not in the absolute value - especially if the System is unloaded == empty flange.
As soon, as payload is attached to the Robot, the measurement Quality is improving intrinsically, since the signal-noise Ratio becomes better - aka there is more Joint torque due to the payload.

DrG

braner331 · September 19, 2019 at 1:41 PM

When estimating powers between surfaces having some fixed geometry, it is far easier to quantify the powers between bended surfaces than between two level surfaces. This is on the grounds that there are no arrangement issues when estimating the power between, state, a circle and a level surface, though two level surfaces must be absolutely level and consummately adjusted at the nuclear level, just as residue free, over an enormous territory for dependable estimations to be made.

hsadeghian · September 26, 2019 at 6:46 AM

Dear All,

I a have the same problem with our kuka iiwa,

The external interaction torque that is shown in the "joint torques" section in teach pendant is not zero even after calibration with the provide program. the order of external interaction on joints is max=1 N.m.

this for sure affects on the measurement of force on the end-effector. Unfortunately it is configuration dependent and the idea of subtraction from initial error sound ineffective.

is there any way to calibrate that?

Regards

braner331 · October 1, 2019 at 4:35 PM

The problem here is calibration and it is normal for any new system. You need to calibrate the force sensor by manually taring its output using the offset value you are getting. You need to find a way to set the no-load output to zero by working with the present value you are getting. Check here for more useful information: https://tacunasystems.com/knowledge-base…easuring-chain/ AND https://tacunasystems.com/knowledge-base…weighing-scale/

iiwa External Force Measurement - High Inaccuracy

Advertising from our partners

Anyware Robotics is hiring!

How to see your Job Posting (search or recruit) here in Robot-Forum.com