Learning of robot arm impedance in operational space using neural networks

Impedance control is one of the most effective control methods for the manipulators in contact with their environments. The characteristic of force and motion control, however, is influenced by a desired impedance of a manipulator's end-effector, which must be designed according to a given task and an environment. The present paper proposes a new method to regulate the impedance of the end-effector through learning of neural networks. The method can regulate not only stiffness and viscosity but also the inertia and virtual trajectory of the end-effector and can realize a smooth transition from free to contact movements by regulating the impedance parameters before a contact.