Comparison of two nonlinear model predictive control strategies with observer-based disturbance compensation for a hydrostatic transmission

Abstract

Based on a decentralised control structure, two fast nonlinear model predictive control (FNMPC) algorithms - a gradient-based FNMPC and a Newton-Raphson-based FNMPC - are investigated and compared for the tracking control of the motor angular velocity of a hydrostatic drive train, which is commercially used in working machines. A flatness-based approach is employed for the tracking control of the normalised bent axis angle of the motor. An unknown leakage volume flow and a resulting load torque are taken into account as lumped disturbances. These disturbances and two unmeasurable state variables - the normalised swashplate angle and the normalised bent axis angle - are estimated by a nonlinear observer. Thereby, a high tracking accuracy of the angular velocity of the motor can be achieved for both FNMPC algorithms. The efficiency of the proposed controllers is demonstrated by experiments.