An Adaptive Robotic System for Doing Pick and Place Operations with Deformable Objects

This paper presents a robot system for performing pick and place operations with deformable objects. The system uses a structured light scanner to capture a point cloud of the object to be grasped. This point cloud is then analyzed to determine a pick and place action. Finally, the determined action is executed by the robot to solve the task. The robotic placement strategy contains several free parameters, which should be chosen in a context-specific manner. To determine these parameters we rely on simulation-based optimization of the individual use cases. The entire system is tested extensively in real world trials. First, the reliability of the grasp is evaluated for 7 different types of pork cuts. Then the validity of the simulation-based optimization of the placement strategy is evaluated for 2 of the most different pork cuts, to show the generality of the overall approach.