Robustness and performance of a vehicle dynamics estimator

To face the nowadays challenges of mobility, the study of vehicle dynamics is of paramount importance since it provides insight on the quantities characterizing vehicle motion, called states. States may be measured precisely and reliably by employing the proper equipment but this is often expensive and very challenging. In order to overcome this obstacle, virtual sensing is used: this technique combines a reduced set of measured data with process models to obtain knowledge on other non-measured quantities. An estimator was built to estimate forces and sideslip angle by using a tire model coupled with an Extended Kalman Filter: it is able to reach a promisingly good performance even in low lateral acceleration scenarios. The aim of this thesis work is to shift the focus of the estimation from the control purposes towards the engineering ones: the target is improving the estimator performance both in terms of dynamic content and behavior of the estimated signal in transient phase, so that the estimate is as close as possible to what the measure of the corresponding quantity would be. First the robustness of the estimator is checked and improved, then the performance of the estimator is tested out and then modified to accommodate new localized measurements, in the attempt of refining the estimate quality.