Tong Liu, Xianren You
This paper addresses the challenges of precise trajectory tracking and proactive error management in hydraulic servo systems subject to inherent nonlinearities, parameter uncertainties, and external disturbances. A novel two-layer intelligent control framework is proposed. Distinct from existing approaches that combine adaptive sliding mode control with generic neural networks, this work introduces an adaptive fast terminal sliding mode controller (AFTSMC) that estimates the lumped disturbance bound online without prior knowledge, supported by a rigorous proof of practical finite-time convergence. In the second layer, a long short-term memory (LSTM) network predicts future tracking errors, with its hyperparameters optimized offline by an improved particle swarm optimization (IPSO) algorithm that incorporates future reference trajectory information. A real-time early warning mechanism is then established based on both the predicted error magnitude and its rate of change. Comparative simulations show that the AFTSMC achieves substantially higher tracking accuracy and robustness than conventional sliding mode control, and the IPSO-optimized LSTM yields significantly lower prediction error and improved goodness-of-fit relative to standard PSO-LSTM, demonstrating the effectiveness of the integrated framework.