This paper proposes an algorithm for robot perception of impedance, which can be used as a fundamental technique for real-time and qualitative perception of physical constraints imposed on the robot's end-effector. This algorithm (1) estimates the impedance that represents the motion-force relation of the end-effector; (2) calculates the uncertainties of the estimates; and (3) detects discontinuous changes in the impedance. The estimated impedance can be used to recognize local dynamic properties of the environment and temporary constraint conditions imposed on the end-effector. The detected discontinuities can be used to segment the manipulated tasks and to recognize the geometric structure of the environment. This method can be implemented in both autonomous and remote-controlled robots because it is designed separately from control methodologies. Results of preliminary experiments are presented.
展开▼
