Learning Deformable Object Models for Mobile Robot Navigation using Depth Cameras and a Manipulation Robot

In this paper, we present our recently developed robotic system that can navigate in environments with deformable objects. To achieve this, we propose techniques to learn models of deformable objects by physical interaction between the robot and the objects. We determine the model parameters by establishing a relation between the applied forces and the corresponding surface deformations as observed with a depth camera. After modeling the objects in a scene, the robot can perform its navigation tasks more efficiently by considering the cost of deformations during path planning. As we demonstrate in real-world experiments, our system is able to estimate appropriate physical parameters that can be used to predict future deformations and exploits this information during path planning.