Load data and arm length/joint position calibration

    Hi all

    Im getting to grips with my krc2 kr210 robot which I use for milling. I want to check a few things and see if I can improve its performance.

    Firstly I've searched Google and cant seem to find the answers so all help here is greatly appreciated.

    I run a 3d scanning and measurement company so have the use of our creaform metrascan for measurement and tracking. I would like to use this tool to accurately map out the arm lengths and axis centres of all the joints and compare this to the values set on the krc2 controller. First questions is therefore where's do I find these values on the controller? And what reference are they in? Ie is the centre of the rotational plane of A1 0,0,0 ? And what format are the values, cartesian or other....

    Next up is , I can measure the weight of the spindle and also determine the cog through scanning the spindle. This is done by scanning the spindle hanging on a rod . Yhe rod naturally hangs vertically and passes through the CoG. By doing this 3 or more times with the spindles weight hanging on the rod in different positions it defines the CoG by the intersecting point of these rod axis.

    I can therefore very accurately map out the joint positions and pay load CoG to approx 10 microns.

    And my hope is inputting these values into the controller, improves the robots performance and accuracy.

    Cheers

    Doug

    Accurate determination of the CG is always good, but the limiting factor for your accuracy will be the robot's inherent limitations.

    Is this an Absolute Accuracy robot?

    NRK's SARCA product does the sort of robot calibration/compensation that you're talking about. It should work with any metrology device that can talk to Spatial Analyzer. The SARCA unit is a small computer that contains a correction model for the robot, generated from a large number of samples, and when queried for a position to move to, will pass back a "corrected" position that will compensate for the robot's measured error.

    It's possible to create your own equivalent in robot code, but you'll need a decent mathematician. You'll need a weighted average of the nearest corrections, and keep in mind that you need to sample in joint space, not just cartesian -- sending the robot to the same location with a different orientation, or even to the same position&orientation but with a different wrist or elbow configuration, will need different corrections.

    However, the big factor here is that, to get maximum accuracy, you need to measure the robot using the actual working payload. Your metrology device may not give you good results unless you add a payload that will bring the total payload up to something very close to the actual working payload you want to use.

    hi thanks for your reply and your suggestion sounds alot like the robot dk approach.

    I firstly want to check how mine is set up as it may have values for these aspects which are incorrect and setting these correctly and specifically for my machine should be set before any other corrective measure is made.

    Does anyone know where these values are held (and set) in the controller so I can firstly compare what we are dealing with?

    Cheers

    Doug

Advertising from our partners