On-line learning of robot arm impedance using neural networks

Impedance control is one of the most effective methods for controlling the interaction between a robotic manipulator and its environments. This method can regulate dynamic properties of the manipulator's end-effector by the mechanical impedance parameters and the desired trajectory. In general, however, it would be difficult to determine them that must be designed according to tasks. In this paper, we propose a new on-line learning method using neural networks to regulate robot impedance while identifying characteristics of task environments by means of four kinds of neural networks: three for the position, velocity and force control of the end-effector; and one for the identification of environments. The validity of the proposed method is demonstrated via a set of computer simulations of a contact task by a multi-joint robotic manipulator.