Reference-tracking Model Predictive Control for highly automated vehicles

The rapid advancement in autonomous vehicle technology have made the field of highly automated vehicles an area of interest within the research community. A particular interest is posed in the design of Motion Controllers for highly automated vehicles, able to provide safety guarantees on the controlled system. The following work proposes a Longitudinal and Lateral (LoLa) Motion Control scheme based on Model Predictive Control (MPC) for highly automated vehicles. The purpose in the design of such controller is to realize a Motion Controller able to faithfully track the provided reference, while being robust and reliable under all circumstances, and providing, under suitable assumptions, mathematical guarantees on the recursive feasibility and stability of the solution. In order to fulfil the desired objective, the Model Predictive Control technique is considered in the control design, allowing to couple the longitudinal and lateral dynamics of the vehicle, while obtaining an optimal control policy with respect to a well-defined cost function. In the following, a walk-through in the field of autonomous driving control solutions and of the designed controller will be provided, involving terminal set design, soft-constrained techniques for ensuring feasibility of the problem, and analytical proofs for guaranteeing robustness and safety of the controller in the linear case. The various design stages will be justified with simulation results on the application of the controller, considering both a linear and non-linear dynamical model of the system. Obtained results will be shown, highlighting the strengths and drawbacks of the proposed approach, showing its applicability in safety-critical applications, such as the one of autonomous driving.