Computer vision algorithm development
Combining camera images, 3D data and machine learning makes it possible to detect and localize objects with variable outlook and size. Input from a depth camera is processed by the computer vision algorithm to detect and localize the objects in 3D.
Design and implementation
A specific gripper is designed and implemented to pick up and hold the objects during movement. It combines suction and clamping. The suction system allows to initially pick up the objects. With it’s foam vacuum cups it can even pick up rough uneven surfaces. The clamping system uses pneumatic cylinders to physically grasp the object and secure it during movement.
The activation of the suction system, position of the sliding and clamping cylinder is controlled using pneumatic valves.
Design, Implementation and integration
A cartesian robot is designed from scratch. 4 axis (X, Y, Z and rotation) are controlled to enable movement of the gripper. The desired gripper position can be calculated after transforming the position from the camera to the robot coordinate system.