Rapid robust trajectory tracking control for oscillatory-base manipulators via enhanced terminal sliding mode and scleronomic Lagrangian mechanics-informed neural networks

This paper proposes a novel trajectory tracking control strategy for a specific type of oscillatory-base manipulator, namely the autoloader, aimed at enhancing its response speed and robustness. Compared to conventional oscillatory-base manipulators, the control design of the autoloader faces greater challenges, primarily due to its stricter demands for rapid convergence and the more complex base oscillations it endures. Developing an accurate dynamic model is essential for achieving rapid control. To address this, a scleronomic Lagrangian mechanics-informed neural network is adopted to model the autoloader’s nonlinear dynamics, which is then integrated into a CTM (computed torque method) - based trajectory tracking framework to enable model linearization. A novel sliding mode reaching law, termed the improved logarithmic-power reaching law, is subsequently proposed. It is combined with a terminal sliding surface to ensure rapid and robust stabilization of the resulting uncertain linear system, with the uncertainty primarily originating from base oscillations. The rapid convergence and robustness of the proposed control strategy are then validated through finite-time stability theory. Finally, both simulation and hardware experiments confirm the effectiveness of the approach, with comparative studies further demonstrating its superiority. • A streamlined trajectory tracking scheme integrates SLMINN for precise dynamics and enhanced TSMC for robust control. • A novel logarithmic-power reaching law is proposed to improve convergence speed and reduce chattering in sliding mode control. • This is the first use of SLMINN in robust model-based control, validated on a hardware platform.