A Deep Koopman-Based Model Predictive Control Method for Valve-Controlled Hydraulic Cylinder Systems

Abstract Hydraulic servo systems are widely applied in construction machinery due to their simple structure and strong bearing capacity. However, considering the nonlinearity and asymmetry in such systems, it is not easy to establish a precise discrete prediction model for the design of the MPC controller, which is a key factor affecting the precision of motion control. To address this issue, this paper proposes a deep Koopman-based model predictive control (MPC) method for valve-controlled asymmetric hydraulic cylinder (VCHC) systems. Significantly, a linear predictor is developed based on the ability of the Koopman operator to lift a nonlinear space to a linear space globally. The simulation results show that the MPC algorithm combined with the Deep Koopman operator has excellent control performance.