Multivariable Nonlinear Data-Driven Control with Application to Autonomous Vehicle Lateral Dynamics

Complex engineering systems are usually described by the interaction of several agents and characterized by highly nonlinear dynamics. Control of multivariable nonlinear systems is a widely explored topic, and many different studies have been presented in the scientific literature. However, most of the existing methods strongly rely upon an accurate model of the system, which is generally costly and/or hard to undertake in practice. In this work, we propose a multivariable extension of the data-driven inversion-based control (D2-IBC) method, where a model of the system is derived from data and considered relevant or not, based only on its weight on the final control performance. This method will prove its effectiveness on a challenging application: the stability control of a four-wheel steering autonomous vehicle.