Design, Simulation and Testing of a linear MPC for lateral dynamic control of a Formula Student Driverless prototype

The automotive world has passed through a large number of evolutionary steps in history, the latter being the development and implementation of autonomous driving capabilities in passenger vehicles, with the objective of making travelling by car more efficient and secure for people. With this awareness, Formula Student, that is one of the most important university student competitions of engineering, introduced the Driverless category in 2017. The aim is to give the opportunity to autonomous driving to develop even faster inside universities, making future engineers able to confront this new technological challenge. This thesis work presents the design and real-time deployment of a Model Predictive Controller (MPC) for vehicle dynamics control in the Formula Student Driverless prototype of Politecnico di Torino. The scope is to propose a simple yet effective approach that can optimally control the yaw dynamics to follow a reference trajectory of a closed loop circuit. To properly design any controller, the system dynamics have to be deeply analysed first. For this reason, a proper simulation model is developed in MATLAB® Simulink, composed by the mathematical forward model of the vehicle, validated with experimental data, and the low level controller already deployed on the real prototype. In this way, an hardware-in-the-loop validation is performed, allowing a full-scale testing of the developed controller in its final configuration before being deployed on the target hardware. A first version of the MPC was developed in MATLAB® Simulink, and proposed for initial controller tuning. The actual controller is written in C++ language to be deployed in real-time on the embedded hardware of the prototype to enhance the code efficiency and runtime. Finally, results of the MPC performance on the real prototype are presented and future developments are addressed.