Vision systems and conveyors

[MartinH]

We are considering a new installation using a Kuka robot, which picks small items off a moving conveyor and sorts them into blister packs.  A vision system would find the items on the conveyor and transmit the position and rotation information to the robot, which would then grip them.

I know this has been done hundreds of times before, but there is something I currently do not understand.

The vision system finds the product on the conveyor, and transmits it to the robot.  However, by the time the robot has gone from its waiting position to the product, surely the product has moved?  Do you therefore have to make an allowance for how far the conveyor (and product) will move before the robot gets there?  If so, how do you calculate this?  When you are testing the installation, you would want to run everything slower, so does the allowance change?  :kopfkratz:

Any information would be greatly appreciated  :supi:

[Jim Tyrer]

hello Martin,
you need to use what is called "line tracking" or (conveyor tracking), in that you get feedback relative to conveyor position from an encoder,

The conveyor position (encoderVal) is trapped when the camera signals "partFound".

When bot starts moving then targetPos is updated by means of offset that you calc as a function of where part was relative to reference position of conveyor and camera when it was found, and where it is now.

             target.Xval = referencePos.Xval OFFSET BY {(part.XrelativeToReference) AND (part.XrelativeToNow)} //how's that for funky code? 

With small parts I would do best to get direction of motion parallel to world X or Y, just make things as easy as possible.
Speed should not be a factor as long as robot can get there soon enough, but give yourself leeway and keep it in your pocket till you are comfortable and then you get to be the hero.

If conveyor encoder is incremental then you will need to (watch for/cope with) the Z pulse, if you're at design stage then you might consider an absolute encoder that can relate to belt position. Every time I've done this I find the designers have dug a hole for me, so if poss. take charge of design now or they may burn your fingers for you.
Best of luck, sport.

[MartinH]

Thanks Jim (I think !!)

I've now re-read the Kuka Function Generator manual, and it seems to make more sense.

Fortunately, the camera we have has an absolute encoder input, so I'll definitely head down this direction from the start.

One thing I cant work out yet is how to connect an encoder up to the Kuka controller.  The Kuka FG manual seems to suggest that an alternative way to do it is to convert the encoder position into 16 digital inputs and then factor an offset.  This seems a little complicated given the processing capability of the controller.

I'm sure that, if I can get this solved, it will open up so many possibilities for us

 

[Jim Tyrer]

wiring an abs. encoder to a set of digital inputs is the only way I've ever done it.
One input for each bit of encoder output and then read all inputs as a word value.

[fishboy]

If you are not trying to do this on a shoestring budget, the KUKA Conveyor.Tech package is what you need. Its a software and hardware bundle. There are resolver and encoder versions. On the resolver version it connects to a channel on the RDW, on the encoder version you get a LP elektronic Can counter module.
The setup is simple, the documentation is good. All you do is offset the conveyor base (base 11 for conv 1) with the camera values for X/Y/Z/A, the Conveyor.Tech software follows the conveyor movements.

Fishboy

[MartinH]

I'm not trying to do it on a budget, but I find it's easier to "do it yourself" than to use "add-on" modules with Kuka robots (although I appreciate this isnt the case for everyone).  Besides, trying to get answers out of my supplier is not simple - I find it quicker to post questions on this and the German forum.

[Murtaza]

We are considering a new installation using a Kuka robot, which picks small items off a moving conveyor and sorts them into blister packs.  A vision system would find the items on the conveyor and transmit the position and rotation information to the robot, which would then grip them.

I know this has been done hundreds of times before, but there is something I currently do not understand.

The vision system finds the product on the conveyor, and transmits it to the robot.  However, by the time the robot has gone from its waiting position to the product, surely the product has moved?  Do you therefore have to make an allowance for how far the conveyor (and product) will move before the robot gets there?  If so, how do you calculate this?  When you are testing the installation, you would want to run everything slower, so does the allowance change?  :kopfkratz:

Any information would be greatly appreciated  :supi:

Dear,

As a last year students we have a provocative project, where we must couple a vision system to KUKA robot.  We are from Belgium, so we can not write so good English.

Ours project don't have a moving conveyor. We must only location the object (coordinates X and Y) and the robot must pick this object up. At these project we use labview vision (software) , the problem which rises is that we don't know how we can get picture from a camera on labview screen, i.e. detection or location of a object.

With  your aid we can complete our project successfully.

Thanks for your quick answer.