Improvement of a non-linear adaptive controller designed for strongly non-linear plants

Hydraulic differential cylinders are strongly non-linear, coupled multivariable electromechanical tools applicable for driving e.g. manipulators or even robots. While hydraulic drives have considerable advantages in comparison with electric ones in several applications their traditional PID control have to cope with the problem of instabilities since the proper setting of the appropriate feedback gains significantly depends on the actual pose of the robot arm and, the temperature of the hydraulic oil drifting in time during operation. To widen their applicability both disturbance rejection based and partial flatness principle based approaches requiring the accurate. model of the system and measurement of the disturbances and their time-derivatives were proposed. For getting rid of needing this accurate set of information an adaptive approach working with a hectic learning phase and an acceptable regime of "tuned" operation was also proposed. In the. present paper considerable non-linear improvement of this control is presented. The "learning" phase shortened and became less hectic, and trajectory, racking became more accurate. No measurement of the disturbance force, and accurate knowledge regarding the parameters of the hydraulic cylinder are needed. The effect of the noise in the measured joint coordinate acceleration is also studied. The capabilities of the method are illustrated via simulation.