• mpc
• simulink
• optimization
• multibody
• fsae
• adams
• casadi
• matlab

Simulators are a key factor in competitions, especially for tuning race car set-ups. In autonomous racing, one important factor is the reliability and the efficiency of the controller that has to manage multiple inputs and multiple outputs without much burdening on the hardware.
In this thesis the goal is to employ an MPC controller with a minimalistic internal model based on the centroidal dynamics to drive the full-fledged vehicle counterpart.
The driven car is instead represented through a very detailed multibody system of the University of Pisa FSAE car (Kerub) modelled in the MSC Adams software package that accurately describes the equations of motion.
The reduced centroidal dynamics of the MPC’s internal model aims at reducing the computation of the prediction horizon in the repetitive iterations. To implement an online controller the CasADi’s optimizer is involved.
The equations of the hyper-realistic model are imported in a Simulink’s block by a ADAMS plug-in.
In the Simulink environment a co-simulation can take place and the two interfaces co-work, exchanging informations at a certain frequency and playing the two mentioned roles. The results are very encouraging, despite the mismatch between the two models that can be handled by the MPC itself.
Virtual test drives show that the centroidal model, enriched with simple results from single-track vehicle dynamics, is able to drive the Kerub almost at the limits of handling.