Magnetic levitation systems (MLSs) exhibit strong nonlinear electromagnetic dynamics and large model variations across operating points, making it difficult for model-dependent controllers such as PID or classical sliding mode control (SMC) to achieve high-precision performance.To address the significant model uncertainty inherent to MLSs, this study proposes an uncertainty-aware control framework that enhances the robustness and efficiency of SMC through a radial basis function neural network (RBF NN).The RBF NN learns and adapts to nonlinearities, operating-point-dependent variations, and lumped disturbances in real time, thereby reducing the burden on the SMC robust term and mitigating chattering.Experimental results on a laboratory MLS demonstrate that the proposed RBF-enhanced SMC reduces the maximum closed-loop sensitivity by 21.1% and decreases sinusoidal tracking error by 80.2% compared with a conventional PID controller.These results confirm that the uncertainty-aware adaptive control significantly improves robustness, accuracy, and stability in magnetic levitation.
Namgung et al. (Thu,) studied this question.