Not linear trajectory

Vlad222 · July 28, 2020 at 4:02 PM

Hi all,

Sometimes robot KUKA KR8 R2100 has strange linear trajectory. I recorded real poses from controller (red points at picture, there is high density of points because of small speed). Green points are program trajectory in src.
It is LIN motion but looks like PTP. How can it be? It's rare case, 99% similar trajectories (with different coordinates) are good (linear).

TCP, load data are right. C_DIS = 0.5 mm but smoothing doesn't affect it because deviation is between points.

Linear commands:
LIN FORWARD(joints[0], error_id) C_DIS
LIN FORWARD(joints[1], error_id) C_DIS
...
Joints are calculated by external computer. Green points are equal to joints after forward kinematics calculation.
Right tool is set to variable $TOOL.

Scale is 1cm (right bottom corner). Smallest distance between program (green) points is ~1 mm. I am not sure but may be rebooting of controller helps for a time.

Fubini · July 28, 2020 at 4:41 PM

Just to be sure about your data: how did you record the positions exactly? Oscilloscope alias Trace? Which channel? $Pos_act? AXIS_act?

Is the robot equipped with absolute accuracy?

How do the recorded positions relate to the joint positions?

Fubini

Loipe · July 28, 2020 at 4:44 PM

try to set the tool with the bas might be not the whole solution but a try worth

Code

  FDAT_ACT         = FDEFAULT
  FDAT_ACT.BASE_NO = Base
  FDAT_ACT.TOOL_NO = Tool
  BAS (#FRAMES )

with the forward function im not sure if it is enough that u just set the tool frame .

because $ACT_TOOL still has the the old Value in maybe it uses the old tool.

using the BAS it sets $TOOL and $ACT_TOOL and other things too.

**panic mode** · July 28, 2020 at 4:57 PM

what is the load?

MOM · July 28, 2020 at 5:04 PM

Hi,

connecting the green points does not look like a linear movement to me!

How did you get the red points?

How do you get the joints[]?

How do you do your forward calculations?

Can you send me some useful data to have a closer look to the problem?

regards

MOM

**panic mode** · July 28, 2020 at 5:18 PM

apparently it is two linear movements

Vlad222 · July 28, 2020 at 5:29 PM

Quote from Fubini

Just to be sure about your data: how did you record the positions exactly? Oscilloscope alias Trace? Which channel? $Pos_act? AXIS_act?
Is the robot equipped with absolute accuracy?
How do the recorded positions relate to the joint positions?
Fubini

$AXIS_ACT and then calculating forward kinematics at external computer. Send to external computer via EthernetKrl.

No absolute accuracy...unfortunately.

We are calculating whole traj at external computer. Then calculate joints and send to KUKA. Green points is FKP of this joints. Red points - FKP of joints received from KUKA by reading $AXIS_ACT.

This problem is very rare. Robot can weld 24h many days and suddenly at one trajectory occurs this problem. I didn't write but robot also slows down during this motion.

Vlad222 · July 28, 2020 at 5:30 PM

Quote from panic mode

apparently it is two linear movements

Green is ideal (program, from src). But red - real trajectory. Why 2 linear?

Lemster68 · July 28, 2020 at 5:36 PM

So is it near a singularity or near the edge of its envelope?

Vlad222 · July 28, 2020 at 5:53 PM

Quote from Lemster68

So is it near a singularity or near the edge of its envelope?

Not near to singularity...and speed is small. This KUKA doesn't has big deviation at singularity.

Vlad222 · July 28, 2020 at 5:54 PM

Quote from panic mode

what is the load?

Torch ~5 kg + load at 3-rd link

MOM · July 28, 2020 at 6:06 PM

hi,

do you like to answer my questions?

MOM · July 28, 2020 at 6:25 PM

Hi,

maybe you just asking for the wrong values

regards

MOM

**panic mode** · July 28, 2020 at 7:25 PM

Quote from Vlad222

Green is ideal (program, from src). But red - real trajectory. Why 2 line

image shows points placed on two different straight segment. short one is on the left, long one is most of the screen.

**SkyeFire** · July 28, 2020 at 7:54 PM

If the problem only happens occasionally, then the first thing to look for is what parts of the process are not consistent. Since you are generating the trajectory from an external computer, what changes from one run to the next? The root cause is most probably something that only occurs rarely.

Not linear trajectory

Advertising from our partners

Anyware Robotics is hiring!

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

Similar Threads

FANUC smoothing with radius (not CNT)

Deviation from linear trajectory

Linear motion near singularity: accuracy vs speed

RSI torque command saturation according to torque prediction

Orientation interpolation (KUKA doesn't use slerp)

Workzone error

Linear and Accurate TCP Speed

Smoothing the trajectory to achieve higher velocity