Vision Requirements for Bin Picking Robots
Automated bin picking can be a benefit to workers by reducing risk of injury from repetitive motion and allowing them to complete more complex tasks. For the system to take the place of a person, it needs to be able to exactly recreate their actions. The robot arm needs to have a range of motion to select product from multiple bins, as well as an appropriate gripper or end effector. Another important requirement is for the robot to identify what part it needs to pick up, which bin to get the product from and where exactly to place it.
To do that, the robot needs a vision system. The vision system would be made up of multiple cameras and software that would allow a robotic arm to recognize and react to objects in the same way a human worker would. This would include recognizing defective products or products that are in an unusual orientation within a bin.
3D cameras provide robots with the ability to “see” in greater detail. Using multiple camera focused on the same view, the images are combined using algorithms that depict an object’s dimensions and texture. This allows the robot to pick one object out of a bin of the same objects. Different algorithms allow for object identification to be made while the object or cameras are in motion, which is ideal for bin picking.
The software and programming associated with the vision systems are crucial to accurate automated picking. Certain cameras and sensors, such as those from IDS Image Development Systems, feature extensions that allow for virtual testing of picks. Virtual testing applications allow programmers to simulate picks to see where errors may occur and provide insight on how to avoid them.
In addition to 3D cameras, visions systems can utilize 2D cameras. These can be used to read bar codes on products or bins to ensure proper picks. 2D cameras also identify borders and edges, allowing the robot to visualize the parameters of the bins.