Input‐Saturated Control of Active Suspension System Using Fuzzy‐Scheduled Estimation of Model Uncertainties and Road Disturbances

ABSTRACT The main challenge in designing a reliable model‐based controller for active suspension systems (ASS) lies in uncertainties in suspension dynamics and unknown road disturbances. In this article, a novel optimization‐based scheme is presented to update a nominal mathematical model, selected initially for the suspension system. In this approach, complementary terms, calculated online from the measured outputs, are appended to the nominal model to compensate for unknown perturbations. The proposed estimation method fuses the information of accelerations and displacements to attenuate the measurement errors, including bias and noise. A fuzzy adaptive solution is presented for effective data fusion to schedule the estimator‐free parameters. The accuracy of the constructed model is confirmed by different experiments performed on a fabricated suspension platform with the Macpherson mechanism. The proposed estimation method showed a significant improvement in accuracy, achieving 21.31% better performance compared to the unknown‐input Kalman filter employing the same acceleration data. In another comparison, the proposed estimation method demonstrated significantly higher accuracy, outperforming the extended state observer and super‐twisting observer by 31.75% and 14%, respectively, in road input estimation. Unlike acceleration‐based approaches, both comparative observers use only displacement data for estimation. To complete the design process of the ASS, a continuous predictive control method is employed using the enriched model to isolate the vehicle body from road irregularities. The stochastic stability of the input‐saturated solution is also analyzed. From the co‐simulation results of Adams/Matlab, the proposed constrained controller demonstrates faster computation times compared to nonlinear model predictive control (NMPC), making it suitable for online implementation. Additionally, the proposed method, which adapts to real conditions by accurately estimating model uncertainties, significantly outperforms the NMPC, achieving a 24.61% improvement in body acceleration.