Robots are going to do many things for us in the future. Before that happens, they need to learn how to handle uneven and soft objects. Picking up a towel like flexible objects is not as straight forward as it seems. Ishikawa Group Laboratory researchers have come up with a robot system capable of manipulating these types of objects.
As they explain:
the towel spreading operation, we extracted the lowest point of the suspended towel as a corner using information from two ToF cameras, and recognized the state of the corner using a machine learning-based method, then we realized the grasping motion of the robot based on the recognition result. After that, a newly developed end-effector (a gripper fixed at both ends and a gripper that moves between them) is used to realize the spreading motion, and the robot arm is used to realize the alignment motion.
The above video shows how this system works. More information is available here.