Control of a cost oriented humanoid robot

Until now the control of the joints was carried out by industrial controllers. These are expensive, heavy and have only limited possibilities for the implementation of advanced control algorithms. Therefore a new hard- and software control concept for the motors and the joints was developed. In order to find appropriate controller parameters the dynamic behavior of the joints is analyzed by means of a nonlinear system identification using a Hammerstein model. The result of the system identification shows that the dynamic behavior of the joints is PT1 element with two nonlinearities, a dead zone and a nonlinear gain. Therefore a “piecewise linear” PI controller - the gain depends from the velocity - will be implemented on a COA processor. Finally an outlook on further works will be described.