Adaptive feedforward control using a gaussian process and a recursive least squares algorithm for a hydraulic axial piston pump

Feedforward control of hydraulic systems is a huge benefit to their performance and has been a research topic for many years focusing on the differential equations of the pump in order to implement pressure control or flow control laws. Considering a servo-hydraulic system which is acutated by an operator, it is inevitable to firstly calculate the required flow to obtain a desired system behavior. A good measure for a suitable pump flow regarding a given input signal by an operator is the resulting pressure drop across the actuated valve. In this paper an adaptive feedforward controller is developed using a gaussian process and a recursive least squares algorithm to calculate the needed flow more accurately resulting in a desired pressure drop repeatedly. The results are then implemented to the hydraulic system and the tracking behavior of the real system is evaluated. This is done using setpoint changes of the operator input. The measurement results for the dynamic actuation of a cylinder are shown and a comparison between the gaussian process and the recursive least squares algorithm is made. With the adaptive feedforward controller the pressure drop can be set more precisely allowing an efficient operation of the overall hydraulic system.