Vision milling system for Kuka Robot

    Hello all,


    I am a proud owner of a robot KUKA LBR4 with LR Controller and KUKA KCP2 Control Panel recently and

    have the idea to turn it into a milling robot witch to integrate to help me in my work.


    I have no experience in tuning robots but I have very good technical knowledge.


    Is there an opportunity to integrate a vision system that scans the details and that a contour or

    area is milled. Because the parts it will process are about 10 varieties, is it possible to be

    scanned and the robot will automatically recognize the given details and perform the desired

    milling according to predefined parameters.


    For example, as seen in the picture, the detail is scanned and the contours are set to be milling

    (for example, letters in white) at a predetermined depth.


    Please give hardware and software suggestions.


    Thank you in advance.


    Best regards.

    An LBR really isn't the best robot for doing milling with, unless you're only milling very soft materials and making very light cuts. The LBR just doesn't have the stiffness or rigidity.


    As far as the vision, maybe? You would need good identifying features on each product, and be able to calibrate them dimensionally. The details would depend entirely on your context -- there are Blob finding tools, circle/edge/line/intersection finders, PatMax... the list is endless. Cognex is generally considered one of the more powerful and flexible vision solutions. Keyence isn't bad either, and somewhat cheaper.

    Thanks for the info really helps me SkyeFire.

    Now I understand that LBR is not suitable for milling of solid materials such as metal and others. It will cut only epoxy resin and plastic.


    With regard to vision systems, can you recommend me what it is like for a Cognex (model/technology) company. With such a vision system, how can I integrate it in my case?


    Thank you in advance.

    There are several in-depth discussions of this in the forum archives.


    In general, you have to choose a communications bus that your robot and vision system both support. This may involve adding hardware or software options. You'll need to check the options available for your robot carefully, then use that list to down-select the vision systems you can choose from. IIRC, the LBR controller supports DeviceNet natively, and adding EIP or ProfiNet to it can be a challenge. The LBR might support RS-232, which is old but works on almost everything.


    You'll also have to determine how many degrees of freedom you need to measure and correct in, which will drive whether you need 2D or 3D vision.


    Then you have to perform the "extrinsic calibration" of the vision system -- basically, calibrating it to a frame of reference that both the vision and robot can share. This can be done several ways, depending on the details of your system configuration (fixed camera, carried camera, etc).


    You have to "teach" the vision system how to identify, find, and measure (in the shared reference frame) the target object(s). Again, this varies greatly depending on the details. Some vision systems give you an "Apple experience", easy to teach but with little control over fine-grained details. Others give you complete control down to the smallest detail, but require you to do most of the work.


    Then, you need to "teach" the vision system and the robot your "perfect" path on your "perfect" part -- you place a part in the optimum position, and program the vision system to treat that part as a zero reference. Then, without moving the part, teach the robot to that part, at that location. The end result, if done correctly, is that the vision system will measure each future part relative to the "perfect" part, and generate offsets based on the differences. And, if the shared reference frame is set up correctly, the robot will be able to apply these offsets and follow the part correctly.


    Lighting is a huge issue. In machine vision, lighting is everything. And the options are nearly infinite.

Advertising from our partners