Design and experimental validation of vehicle dynamics estimators and trajectory tracking model predictive controller for scaled fully autonomous vehicles

The progress of self-driving cars widely depends on the development of precise navigation systems and effective motion controllers that allow to track the desired trajectories provided by the motion planning stage. Both these critical aspects are addressed in this work: the enhancement of localization accuracy by using sensor fusion and estimation techniques and the design and experimental validation of trajectory tracking controllers for autonomous driving. In particular, the first part of the thesis focuses on the design of Extended Kalman Filters (EKFs) that, by integrating measurements of multiple sensors such as Lidar, Inertial Measurement Unit (IMU) and encoder, allow to overcome the issues of splikes in Lidar-based pose measurements. These filters not only mitigate the impact of outliers in the Lidar measurements but also allow to estimate vehicle dynamics state variables that are not directly measured, e.g. vehicle side-slip angle. The improved localization and vehicle dynamics estimators are crucial for ensuring both reliability and accuracy of autonomous vehicles in diverse operational scenarios. The second part of the work delves into the motion control stage which represents a fundamental module of an autonomous vehicles software stack, aiming to follow accurately the reference path by ensuring vehicle stability and robustness of control system performance. Specifically, two control strategies are presented: a static state feedback controller with optimization and a model predictive controller. All estimators and controllers have been firstly tested by using closed-loop simulations and then experimental validated employing a scaled fully autonomous vehicle prototype.