Pivoting manipulation is an effective method for carrying large and heavy objects which have high friction between the floor. In this paper, we propose the motion generation and control method for pivoting manipulation adaptive to the object state and contact force. The robot estimates the actual object motion and the required manipulation force online based on the kinematic and physical sensor feedback, and modifies the object path and the whole-body posture for pivoting manipulation. By this pivoting system, a robot can pivot the object without the knowledge of the object mass and friction parameter. Finally, we confirm that the proposed pivoting system enables a robot to acquire the adaptability to the variation of the object weight and the error of the object motion by the experiment, in which the life-sized humanoid robot carries the large furniture.
