Human-Robot Collaborative Handling of Highly Deformable Materials

Abstract: 

Robotic manipulation of highly deformable materials like cloth is a challenging problem due to the high dimensionality and the need for multiple grasped points to fully manipulate the material. Previous methods focused on developing detailed models for the material representation and its interaction with the robot in order to generate a motion plan. In this paper, we expand on our previous work in co-robotic manipulation by considering the collaborative transport of a deformable material by human and a mobile dual-arm robot. Our approach is to first determine optimal position and velocity setpoints from human pose information; the mobile dual-arm robot then attempts to follow these setpoints while satisfying other constraints. The goal is to stably and efficiently transport the material to a desired state. As the human changes his/her pose, the robot determines the relative optimal pose of its end effectors and uses a feedback control law to determine desired end effector velocities. The control signal to the actuators that meets the desired end effector velocities is then calculated according a constrained quadratic program. We evaluate the controller in simulation using the Bullet physics engine to approximate the cloth behavior.

Reference:
D. Kruse, R.J. Radke, J.T. Wen (2017). Human-Robot Collaborative Handling of Highly Deformable Materials.

American Control Conference, May, Seattle, WA 2017.

Publication Type: 
Conference Articles