Neural-Network Adaptive Output-Feedback Saturation Control for Uncertain Active Suspension Systems.

Area of Science:

• Control Systems Engineering
• Automotive Engineering
• Artificial Intelligence

Background:

• Active suspension systems are crucial for vehicle dynamics.
• Unknown parameters and unmeasured states pose challenges in control design.
• Neural networks (NNs) offer powerful tools for approximating complex dynamics.

Purpose of the Study:

• To develop an adaptive NN output-feedback control for a quarter-car active suspension.
• To address unknown sprung mass and suspension stiffness.
• To estimate unmeasured states using an NN state observer.

Main Methods:

• Adaptive backstepping control design.
• Neural networks (NNs) for approximating nonlinear dynamics.
• NN state observer for estimating unmeasured states.
• Command filter method to handle state estimation.
• Auxiliary system to compensate for input saturation.

Main Results:

• All system variables are proven to be bounded.
• Guaranteed ride comfort, ride safety, and suspension space limitations.
• Demonstrated effectiveness through computer simulations and comparative analysis.

Conclusions:

• The proposed observer-based NN output-feedback control is effective for active suspension systems.
• The control strategy successfully handles unknown dynamics and unmeasured states.
• The method ensures performance criteria including ride comfort and safety.