Model-Data Hybrid Driven Control of Hydraulic Euler–Lagrange Systems
Zhikai Yao, Xianglong Liang, Shuping Wang, Jianyong Yao
Abstract
As hydraulic Euler–Lagrange systems evolve to encompass high-dimension and intricate structures, the formulation of the mechanical dynamics model becomes impractical, even though the hydraulic dynamic information remains accessible. In this article, we make the first attempt to establish model-data hybrid driven controller for hydraulic Euler–Lagrange systems. The proposed strategy commences with the formulation of model-based nonlinear robust controller, followed by the introduction of actor–critic reinforcement learning to effectively tackle uncertainties associated with unknown mechanical dynamics. Employing Lyapunov analysis, we have proven the asymptotic stability of the model-based nonlinear robust controller, along with the convergence and (sub)optimality of the introduced actor–critic reinforcement learning. Building upon this foundation, the newly devised hybrid-driven controller still guarantees asymptotic stability through the reconstruction of the error auxiliary function. We validated the efficacy of the proposed strategy on a well-established six-degree-of-freedom hydraulic manipulator and conducted a comparative analysis with the purely data-driven approach to demonstrate the superior performance of the developed controller. The integration of model-based and data-driven terms in our strategy signifies notable advancements in the control paradigm for hydraulic Euler–Lagrange systems, and the conceptual control framework also holds the potential to address control challenges within diverse and intricate industrial systems.