Dual-Channel Uncertainty Estimation-Based Residual-Augmented Preview MPC for Active Suspension Systems

Ride comfort and safety of autonomous vehicles are strongly influenced by active suspension systems (ASS). However, modeling uncertainties caused by unmodeled dynamics and parameter variations may degrade prediction accuracy and limit control performance. To address this issue, a preview uncertainty-aware residual-compensated model predictive control (PURC-MPC) framework is proposed in this article. Specifically, a residual-compensated linear state-space model is constructed to enhance prediction accuracy. Meanwhile, a dual-channel online uncertainty estimator is developed to quantify residual reliability and constraint-related risk, and is incorporated into the preview MPC formulation to modulate residual usage and enforce safety margins. Simulation and experimental validation on a Quanser quarter-vehicle ASS (QV-ASS) bench is conducted under speed-bump, sinusoidal, and random ISO Class-B road excitations. Results show that, compared with the passive suspension system (PSS), the root-mean-square error of sprung mass acceleration (SMA) achieved by PURC-MPC is reduced by 77.41%, 52.52%, and 56.92%, respectively. These results indicate that PURC-MPC provides an effective and implementable solution for robust comfort-oriented suspension control under uncertain dynamics.