Model free control design for a semi-active suspension of a passenger car

Comfort and road handling in passenger cars can be improved by including a semi-active suspension system, controlled by online varying the damping based on measurements of the car motions. The selection of an appropriate control structure is crucial, since it determines the complexity of the control design and parameter tuning process. Model based control requires the identification of a physical or black-box model of the car and semi-active suspension dynamics, which is not straightforward and time-consuming, especially when these dynamics are highly nonlinear and strongly coupled. Model free controllers, such as neural networks, can be hard to train or intuitively fine tune manually since its control parameters have no physical meaning. This paper presents a flexible model free control structure based on physical insights in the car and semi-active suspension dynamics used to linearize and decouple the system, such that decentralized linear feedback control can be applied. The resulting controller can be tuned online intuitively using a limited number of physically interpretable parameters.