Application of a hybrid controller with non-contact impedance to a humanoid robot

This paper addresses the design issues of a hybrid neural network-based PID control with grey prediction and non-contact impedance as applied to a humanoid robot named ISAC. ISAC uses two six degree-of freedom manipulators called soft arms which are actuated by pneumatic artificial muscles. High nonlinearities such as hysteresis and unknown dynamics of these muscles make arm control a challenging task. ISAC was designed to assist physically challenged, therefore it has to fulfill safety precautions concerning its movements. To evaluate the proposed controller, it was tested for human-like grasping motions while guaranteeing the safe interaction with humans. The controller consists of two sub-controllers, an open-loop neural network-based controller which is managed by non-contact impedance and powered by grey prediction.