Finding a hole

Student

Hi! I am working on an application where the cobot has to take a gauge with the gripper and then go to a hole in an engine to measure. I am using to find the hole a camera from Cognex. The problem is that when I send the coordinates to the cobot and moves to the hole to measure, I loose some precision, so sometimes it is impossible to enter into the hole. 

Do you know how can I improve the application finding the hole? I know the best solution is to use a force torque sensor (with spiral search or so) but I don't have any actually...

Thanks!

matthewd92

@Student So here are a couple of examples of using a pin to gauge a hole.  The first image is of a standard type pin gauge, where there is a nice square corner which you as a human can still easily handle.  The second shows the gauge having some lead-in, quite exaggerated in this example.





As you can imagine its much easier to get the second one started in the hole.  By using the Force wizard you can place a waypoint inside of a force node and allow the robot to float in the XY directions if you are for example moving in the Z direction to insert the pin.



As far as how to implement the spiral search in polyscope refer to this post.  The difference is you will need to place into a background thread a feature to find the forces on the robot

Thread
  forceOnArm = get_tcp_force() #This returns a list of forces in X,Y,Z, Rx, Ry, Rz on the TCP
  Fx = forceOnArm[0]
  Fy = forceOnArm[1]
  Fz = forceOnArm[2]
  sync()

 

You could the just substitute this into program that @Tyler_Berryman produced.  The thread where the idea Tyler did came from is located here

Hope this helps

matthewd92

You could use the spiral search with the built-in force monitoring. It's not as precise as an external FT sensor but may work. 

How repeatable/precise is is the measurement coming from the camera?  Understand the robot only has a precision of .004"/0.1mm so if your hole tolerance is less than that the robot will have a hard time anyways. 

Another option without seeing your gauge is to provide some lead-in on the edge of the gauge and then allow the robot to float in the XY directions while you move the gauge in the Z direction allowing the gauge to find its own way into the hole.  You would use the force wizard and a frame motion type to achieve this. 

Student

@matthewd92 said:

You could use the spiral search with the built-in force monitoring. It's not as precise as an external FT sensor but may work. 

How repeatable/precise is is the measurement coming from the camera?  Understand the robot only has a precision of .004"/0.1mm so if your hole tolerance is less than that the robot will have a hard time anyways. 

Another option without seeing your gauge is to provide some lead-in on the edge of the gauge and then allow the robot to float in the XY directions while you move the gauge in the Z direction allowing the gauge to find its own way into the hole.  You would use the force wizard and a frame motion type to achieve this. 

@matthewd92 Thanks for your response.

I don't know exactly what you mean in your third paragraph. And how can I implement the spiral search in Polyscope?

Thanks!

Student

Thanks @matthewd92 !

Just one more question. Can I move in X and Y axis while applying a force in the Z axis? I mean, I want to do the spiral search keeping the force in Z axis until the cobot finds the hole, so the Z axis will change.

matthewd92

@student. You're more than welcome.

To answer your other question, yes you can do that. Just use the Forze wizard and make the z-axis compliant using the tool frame, then give it the amount of force to use. Also, one tip, reduce the amount of speed that it uses from the default amount down to something like 25 mm/sec and then work up in speed until it does what you want. 

Student

Thanks again @matthewd92

I have implement the script and edited and now I can use it. But I have a big problem, I need to measure different holes of a block (from an engine) but to do the spiral search I need to contact the gauge and the block, so doing the spiral search the gauge scratches the block while finding the hole. Is it other way to do it without contact?

matthewd92

If you have chamfered the edges of the pin you should be able to get into the hole without hitting the surface.  The spiral should only be moving a small amount. 

Its really hard hard to know what to tell you without being able to see the actual application. 

Student

What I am doing is a peg-in-hole application using an UR10 + Cognex camera to measure the depth of different holes in a block (part of an engine). So, I need to find the hole and the position with the camera to send the coordinates to the cobot, then the cobot moves there and introduces the gauge into the hole to measure. Finally the cobot extracts the gauge and goes to the next hole. The problem is that I loose some precision when I transform from pixels to cartesian coordinates (in mm)... And I can't chamfer the edges of the gauge.

So I need to find exactly the center of the hole but the spiral search sractches the block... @matthewd92

Loïc

I don't understand why you can't chamfer the edges of the gauge, this seems to be a pretty good option.

To get a better precision with your optic system without changing your camera, you can try to reduce the field of view (if it's possible...) but you'll probably have to change/add a lense to your camera.

Another option could be using a force sensor like the FT300. You'll still scratch the block but as you'll get a better precision with this kind of sensor you'll be able to apply less pression.

matthewd92

@Student I agree with @Loïc if you aren't able to chamfer the edge of the gauge you will either need to increase the precision you are using to find the hole or you will need to use an FT sensor. I can say we are using a camera to find a 4 mm hole in a part and then place that hole onto a peg that's about 3.8 mm square.  We do not use any type of search method to do this either.  To achieve the precision that we have the hole taking up about 70% of the FoV of the camera, it's literally placed within 75 mm of the part where the hole is. 

Student

Finally I did what I wanted. I am using a camera for guided vision and also a FT sensor. Now I can find precisely the hole and enter with the gague, also I have "spiral search" function because of the camera precision with the coordinates

Thanks to everyone!

Karine_Simard

Congrats! Can you post a video of your program?

Student

@Karine_Simard Not yet, because of the company...

Student

Here you have a little example, @Karine_Simard

https://streamable.com/bgwtl

Karine_Simard

Thanks! Nice work!