The goal of this project is to model the behavior of objects grasped with a flexible vacuum gripper. Since the vacuum gripper uses a flexible suction cup to grasp objects, heavy objects will cause the cup to bend and put the grasped object at an angle offset to the gripper pose. It is a non-trivial task to determine the pose of the grasped object given the end-effector pose without a suitable dynamics model for the vacuum gripper - rigid object system. To explore this problem, we created, fit, and tested a candidate model for the dynamics of a vacuum gripper - rigid object system and demonstrated its feasibility to predict object pose given gripper pose on a number of tests. A proper model will help make planning and control algorithms take into account the non rigid behavior.. The end goal of this project is to determine a workable model of vacuum gripper dynamics and demonstrate the feasibility of our proposed model on a number of tasks.

Example: Imagine picking up a heavy box with a flexible vacuum gripper attached to a robot-arm like the Sawyer Arm. Manipulating the object by actuating the robot will induce unpredictable motion in the heavy box as the box causes the suction cup on the vacuum gripper to deform under its weight. Thus, even relatively simple pick-and-place tasks require additional sensing and more complex system design when using vacuum-grippers. A suitable model of vacuum-gripper - rigid object system would allow for more efficient planning and manipulation with less sensing required potentially cutting costs while increasing accuracy.

There are countless real-world applications using vacuum grippers. In industry, vacuum grippers are used in pick-and-place tasks, palletization in distribution centers, part assembly tasks in manufacturing facilities, and in any use case where traditional grippers/end-effectors struggle like when components are of different materials and shapes. Due to the widespread use of vacuum grippers, eliminating camera-based sensing by relying on a suitable model for object-vacuum gripper interactions could potentially improve system reliability and cut down on sensing costs in real-world applications. Additionally, reliable manipulation of objects with vacuum grippers could allow for more intelligent planning and less costly actuations; ideally, equipped with a reliable dynamics model of the object-vacuum gripper system, a robotic arm could use its surroundings as leverage points to move the object to the desired pose in a more economical fashion. Reliable manipulation of vacuum-gripped objects provides many exciting real-world use cases.