Constrained State and Input Estimation for a MacPherson Suspension Using the Unscented Kalman Filter and a 3D Multibody Model

Introduction Research on control strategies for semi-active and active suspensions of vehicles has been carried out extensively in the last two decades. 1D and 2D analytical models, like the quarter-car model, are commonly employed for suspension control synthesis and analysis. 3D analytical models of both the suspension and the full vehicle have also been employed more recently. The skyhook, groundhook or hybrid control algorithms are usual model–free control algorithms for semi-active suspensions (SAS) that are tuned using the aforementioned models [1]. More advanced control techniques such as optimal control, adaptive control, pole location control, fuzzy control and sliding mode have also been applied to the control of semiand active suspensions [2]. All these control strategies assume that velocities (for example the one of the sprung mass or the one of the damper) can be measured. However because such sensors are not widely available in practice, instead the velocities have to be computed both from position measurements by derivation or from acceleration measurements by integration, which is a challenging task. For example position sensors are not suitable for measurements above ~20 Hz while acceleration sensors are not suitable for measurements under ~5 HZ [3]. The aforementioned reasons justify the use of model-based observers to compute these quantities and make them available to the controller as virtual sensors [4].