Adaptive Nonlinear PID Control of DC Motor Position Using a Polynomial Fuzzy Long Short-Term Memory Neural Network

This paper presents an adaptive nonlinear PID control approach enhanced by a polynomial fuzzy long short-term memory (LSTM) neural network-based system identification, addressing the limitations of traditional PID controllers in nonlinear systems. The proposed PFLSTM-ANPID integrates polynomial fuzzy modeling with LSTM-based learning, dynamically adjusting PID gains in real-time without requiring prior knowledge of system dynamics. This innovative adaptive mechanism enhances control precision, disturbance rejection, and real-time adaptability in complex, time-varying environments. Simulation studies on a highly nonlinear system demonstrate that the proposed controller significantly outperforms traditional PID, achieving 80.51% reduction in Root Mean Square Error (RMSE), 81.28% reduction in Integral Absolute Error (IAE), and 85.39% reduction in Integral Time Absolute Error (ITAE). Additionally, it surpasses advanced adaptive controllers, including SISO-CFMFAC variants, attaining the lowest IAE of 98.502. Experimental validation on a DC motor position control task further confirms the controller’s superiority, showing a 47.85% improvement in RMSE, 68.91% in IAE and a 78.57% improvement in ITAE compared to traditional PID, and outperforming a fuzzy neural network-based adaptive PID (FNN-APID) controller across all metrics. The proposed approach offers a robust, scalable, and practical solution for industrial applications where system parameters fluctuate significantly over time, representing a meaningful advancement in adaptive control techniques.