Neuromorphic control for robotic manipulators-position, force and impact control

Neural network applications for robotic motion control are considered. The `neural servocontroller´ is based on a neural network which has integrated time-delay elements so that it can learn the nonlinear dynamical system The controller is applicable to position and force control of robotic manipulators. Because they have strong nonlinearity and are high-speed phenomena, it is difficult to sense collisions and to control robotic manipulators suffering collisions. Therefore, the neural servocontroller is effective against collision phenomena. A Herz-type model with an energy loss parameter was adopted to express the impact force between the manipulator and unknown objects. Simulation results show that the proposed neural servocontroller can sense collision phenomena and control robotic manipulators suffering collision