Gripper
Tasks 1, 2 and 3 could be completed using the existing XArm7 parallel jaw gripper, however task 4 required a custom design. The approach was to design a gripper capable of executing task 4 without interfering with previous success for the other tasks. Thus the gripper had 3 design constraints: 1) sticky surface to pick up bill; 2) significant compliance to overcome original gripper movement in z direction when opening and closing; 3) leaving original fingers exposed. The final design - as shown in the image to the left - consists of two external fingers mounted onto the original gripper. Fingers are composed of an outer layer of silicone (shoreness 10A cast into 3-4mm rectangular sheet) and inner layer of sponge (traditional dish-washing sponge in two pieces). They are then mounted using silicon bands in two locations.
General Description of Software
While each item had unique steps in how to be grabbed, the general structure of the software for all four tasks was very similar. The first step was always to locate both the object and the place to put the object. If the target (either the object or the place to put the object) was not in frame, the camera (which was attached to the wrist) was moved up 0.1 m to expand the field of view. Once the target was in frame, the camera was moved such that the target was directly in the center of the frame. This strategy allowed for better measurements on the location of the target's center in the edge case where only part of the object was in frame (which caused inaccurate measurements of the location of the center). The pixel coordinates for the center of the target were converted to real-world coordinates for the XArm7 to perform Pick-and-Place on.
The GitHub repository for our code can be found here .