Pick and place applications have a common problem: how do you pick unusually shaped parts from a container? This becomes even more complex in low volume, high mix applications.
Even for modern vision systems, randomly arranged parts in a container can present challenges… or do they?
In the video above we speak with Ryan Guthrie, executive VP of TM Robotics, about how modern machine vision systems can identify parts without using CAD data for reference, enabling more accurate and flexible part picking, with the TSVision3D software.
“Toshiba Machine has developed their own in-house software for 3D bin picking and one of the key features of our software is that you do not need a CAD file to learn a part,” said Guthrie. “You’re able to take a part, and present it to our vision system in multiple orientations to create a point cloud of that part. The model is then stored in the system and with that we’re able to pick the parts based on that information within minutes.”
Parts with complex shapes and the density of parts can sometimes present issues for a gripper. With the TSVision3D software, a part is analyzed by the vision system to determine optimum angles for gripping and the order in which parts are picked up.
“We can set interference areas in the image and the model to make sure a part is clear so we can grab the part cleanly,” Guthrie explained. “Our system also learns the tool and learns the box, so we can do collision avoidance, and we do that before we even send coordinates to the robot.”
For high mix applications, the software can be used to identify specific parts in a collection of other parts within a single container and only pick those parts. It can even identify different containers.
The software is currently optimized for 5kg and lighter types of parts and smaller containers, for electronics assembly and similar applications, Guthrie explained.
For more information about how the TSVision3D software works, watch the video above and visit the TM Robotics website.