Simulator-enhanced Estimation Loop (SeEL): vehicle model optimisation for state estimation

A reliable estimation of key vehicle states is crucial for enhancing safety in modern automotive control systems. In this paper the Simulator-enhanced Estimation Loop (SeEL) is presented–a gradient descent-based algorithm designed to optimise the parameters of vehicle model, so as to best match the behaviour of a reference vehicle (real vehicle data or any other target vehicle model). A comprehensive state-space formulation of a 5-degree-of-freedom (DOFs) model is presented. The algorithm is then described, then it is validated through simulations, at first, and then on experimental data obtained with a real vehicle. When tested against (more challenging) experimental data, the SeEL demonstrates the ability to adjust vehicle parameters effectively and force the model to mimic the behaviour of the reference vehicle.