I use a gripper to insert and remove samples in the rack module by setting a fixed position for insertion and removal. However, the rack module sometimes shifts slightly, causing the gripper to move to the preset position but fail to insert the sample accurately. Is there a way to enable the gripper to locate the correct insertion point even if the rack module moves? If so, can you provide a detailed tutorial on how to achieve this?
PS: I use RTDE python interface, so I need to use code to control the robot/gripper not TP.
a little updated : we have wrist camera in the robot
Repeatability is one of the largest hurdles to overcome when working with robots. In my opinion, the right answer is to ensure the rack doesn’t move. Design some better fixturing, use clamps to lock the rack down, etc. I work for a company that does a lot of robotic integrations and while everyone likes to reach for some software solution, the simplest answer is usually to fix the upstream problem.
2 Likes
@eric.feldmann Thank you for the reply, But the rack module is installed on a machine with wheels, and sometimes people bump into the machine and cause it to move. Therefore, the rack module cannot be fixed, and sometimes the slight movement of the wheel will cause the insertion position to be inaccurate. Is there a method to determine whether the rack module has moved after the gripper reaches the specified position? This would verify if the sample can be inserted correctly. If the sample cannot be inserted, the gripper could issue a warning, allowing for manual adjustment of the position.
I’ll stay on my soap box only a little longer, I promise. Many of our integrations involve moving carts in and out of a cell on wheels. We have entire hundreds-of-pounds gripper assemblies that change out via a rolling cart. No programming fanciness. I promise you, you CAN accurately locate wheeled carts that can tolerate being bumped by an operator.
(I’ll leave a link to the type of mechanism I’m talking about here: https://www.youtube.com/watch?v=HiOiBh_ry9I We use this sort of thing to dock mobile robots/workstations for accurate location all the time.)
That said, if all you’re trying to accomplish is tell the operator “hey buddy I can’t insert this part” then sure, just go to your Waypoint Node and in the bottom right there should be a little toggle slider that says “Add Until.” Click that and you can choose Tool Contact. Then choose “Add Action” and do whatever you want. That way if your Tool makes contact (ie doesn’t go in the hole) you can react.
I understand you’re driving this with RTDE and python, which in my opinion adds another layer of complexity. However, you can still just whip up a quick dummy program using the teach pendent, save it, and open its script to see the URScript that’s running behind the scenes. Then you just have to make your python write those lines instead.
I would argue that you almost have this behavior already, seeing as if the robot can’t insert the sample into the hole it likely throws a protective stop. (That’s a little speculation on my part, maybe it just crushes the sample and moves on. I have no idea what your parts look like lol.)
1 Like
I’m working on a solution to the alignment problem for mobile robots.
Here is a short video with an early demonstration with the high accuracy aligner (Selfie Aligner).
Transistor Test: Selfie Aligner - Transistor Test
I’ve had to solve this in an older factory and low budget for the project.
Use a laser sensor on your EOAT that can look straight down. Set points so that the sensor is moved past the edge of the tray in “X” and then in “Y”. Save the actual TCP in each case (when the sensor triggers) and then the X and Y within. Do this with the tray in a known position (pushed back and left for instance, This is the position you will also teach your points in). Find the delta between the old and new X, Y and add the offsets to each move.
It’s easier than it sounds and the accuracy is determined by the speed you travel for the measurements.
Thank you for your reply, but I am not sure we have the laser sensor, but we have wrist camera.
Ychen3,
They are only $270 at Mcmaster and can be attached to the End of arm electrical connector. A lot cheaper than the camera and much easier to troubleshoot.