Massimiliano Baldinu
ARTIFICIAL INTELLIGENCE TECHNIQUES FOR PATH TRACKING OF AUTONOMOUS SCALED VEHICLE.
Rel. Mauro Velardocchia, Aldo Sorniotti, Antonio Tota, Luca Dimauro, Matteo Depalo. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Meccanica, 2026
|
|
PDF (Tesi_di_laurea)
- Tesi
Accesso limitato a: Solo utenti staff fino al 1 Ottobre 2029 (data di embargo). Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (7MB) |
Abstract
Autonomous vehicle control requires precise trajectory tracking, especially when operating near the limits of vehicle handling. Traditional Model Predictive Control (MPC) strategies generally rely on simplified analytical representations, such as kinematic or dynamic bicycle models. While effective under nominal conditions, these models struggle to accurately capture complex, non-linear physical phenomena like tire saturation and friction loss. To overcome these limitations, this thesis presents the development and validation of a data-driven Neural Network Model Predictive Controller (NNMPC), designed for the Quanser QCar 2 experimental platform. The proposed architecture replaces the standard mathematical plant with a Long Short-Term Memory (LSTM) recurrent neural network.
By learning directly from experimental data, the LSTM acts as the internal predictive model, forecasting the lateral dynamics, that is, yaw rate and lateral acceleration
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Corso di laurea
Classe di laurea
URI
![]() |
Modifica (riservato agli operatori) |
