Following human guidance to cooperatively carry a large object

Abstract

Carrying a large object like a table is a task that cannot be solved by a single robot or a single human, but that requires two workers, for example one human and one robot. For this human-robot cooperation, the robot must perceive the human and synchronize with its motion. It also must perceive the object to carry. In this paper, we present an approach that uses arm compliance to follow the human guidance on a fast time scale and moves the robot base to restore a nominal position for the arms. For perceiving the object, we acquire a model of it using an RGB- D camera and match this model with the current measurements. This real-time object pose estimate is suitable for approaching and grasping it, as well as for the detection of object lifting and lowering to the ground again. We evaluate our approach in lab experiments using a robot that has an anthropomorphic upper body and an omnidirectional base. We also report on the successful public demonstration of our approach in the ©Home league at RoboCup 2011.