This thesis presents the implementation, tuning and experimental validation of a real-time control strategy to enhance the Lateral Vehicle Motion Control (LVMC) of a SUV equipped with front and rear Steer-by-Wire (SBW) and Brake-by-Wire (BBW) systems. The core of the proposed approach is an incremental linear Model Predictive Control (MPC) strategy, integrated within a hierarchical control architecture composed of a Reference Generation, the MPC itself and a Yaw Moment Allocator. The MPC computes the optimal incremental front and rear steering angles and yaw moment, generated through the braking system, to track a reference yaw rate while minimizing actuator effort. The innovative contribution of this work is the real-time implementation and on-vehicle validation of the controller. Starting from a controller developed and tested in a software environment using Simulink and a Driver-in-the-Loop (DiL) simulator, the MPC is adapted for execution on a physical vehicle using the dSPACE platform, which enables data acquisition, parameter tuning and on-vehicle execution. The implementation process includes the definition of the predictive model and the cost function, followed by the controller validation. The process begins with open-loop experiments to identify the parameters for the predictive model, such as cornering stiffness and actuator dynamics. The identified model is then experimentally validated and the controller is implemented on the physical vehicle. Subsequently, an empirical incremental tuning process is applied to determine the cost function parameters, using steering pad, double lane change and steering step maneuvers to excite both static and dynamic vehicle behavior. Finally, the controller is tested on the vehicle to assess its performance and robustness. Results show improved lateral stability, handling and tracking accuracy compared to the baseline vehicle, demonstrating the feasibility of a real-time MPC, bridging the gap between simulation-based design and real-world implementation.
Questa tesi presenta la progettazione, la taratura e la validazione sperimentale di una strategia di controllo in tempo reale volta a migliorare il Lateral Vehicle Motion Control (LVMC) di un SUV dotato di sistemi Steer-by-Wire (SBW), anteriore e posteriore, e di un sistema Brake-by-Wire (BBW). Il cuore del controllore è un Model Predictive Control (MPC) lineare e incrementale, inserito in un’architettura di controllo gerarchica composta da una Reference Generation, dall’MPC stesso e da uno Yaw Moment Allocator. L’MPC calcola gli incrementi ottimali degli angoli di sterzo, anteriore e posteriore, e lo yaw moment, generato dal sistema frenante, al fine di seguire uno yaw rate di riferimento minimizzando l’utilizzo degli attuatori. Il contributo innovativo consiste nell’implementazione in tempo reale e nella validazione del controllore sul veicolo reale. Partendo da un MPC sviluppato e testato in ambiente software tramite Simulink e un simulatore Driver-in-the-loop (DiL), il controllore è stato adattato per l’esecuzione a bordo di un veicolo fisico. La piattaforma dSPACE ha permesso l’acquisizione dei dati, la taratura dei parametri e l’esecuzione del codice sul veicolo. Il processo di implementazione ha incluso la definizione del modello predittivo e della funzione di costo, seguiti dalla validazione. Per identificare i parametri del modello, come la cornering stiffness e la dinamica degli attuatori, sono state condotte prove in open-loop. Il modello ottenuto è stato poi validato sperimentalmente e il controllore implementato sul veicolo. Successivamente, è stata condotta la taratura dei pesi della funzione di costo mediante un approccio empirico e incrementale, utilizzando manovre come steering pad, double lane change e steering step, per sollecitare sia il comportamento statico che dinamico del veicolo. Infine, il controllore è stato testato sul veicolo per valutarne prestazioni e robustezza. I risultati mostrano un miglioramento della stabilità laterale, della manovrabilità e della precisione di tracciatura rispetto al veicolo base, dimostrando la realizzabilità di un MPC in tempo reale, colmando il divario tra progettazione software e implementazione reale.
Implementation, tuning and experimental validation of a model predictive controller for lateral vehicle motion control
Wu, Sofia
2024/2025
Abstract
This thesis presents the implementation, tuning and experimental validation of a real-time control strategy to enhance the Lateral Vehicle Motion Control (LVMC) of a SUV equipped with front and rear Steer-by-Wire (SBW) and Brake-by-Wire (BBW) systems. The core of the proposed approach is an incremental linear Model Predictive Control (MPC) strategy, integrated within a hierarchical control architecture composed of a Reference Generation, the MPC itself and a Yaw Moment Allocator. The MPC computes the optimal incremental front and rear steering angles and yaw moment, generated through the braking system, to track a reference yaw rate while minimizing actuator effort. The innovative contribution of this work is the real-time implementation and on-vehicle validation of the controller. Starting from a controller developed and tested in a software environment using Simulink and a Driver-in-the-Loop (DiL) simulator, the MPC is adapted for execution on a physical vehicle using the dSPACE platform, which enables data acquisition, parameter tuning and on-vehicle execution. The implementation process includes the definition of the predictive model and the cost function, followed by the controller validation. The process begins with open-loop experiments to identify the parameters for the predictive model, such as cornering stiffness and actuator dynamics. The identified model is then experimentally validated and the controller is implemented on the physical vehicle. Subsequently, an empirical incremental tuning process is applied to determine the cost function parameters, using steering pad, double lane change and steering step maneuvers to excite both static and dynamic vehicle behavior. Finally, the controller is tested on the vehicle to assess its performance and robustness. Results show improved lateral stability, handling and tracking accuracy compared to the baseline vehicle, demonstrating the feasibility of a real-time MPC, bridging the gap between simulation-based design and real-world implementation.| File | Dimensione | Formato | |
|---|---|---|---|
|
2025_12_Wu_Executive_Summary.pdf
non accessibile
Descrizione: Executive Summary
Dimensione
5.16 MB
Formato
Adobe PDF
|
5.16 MB | Adobe PDF | Visualizza/Apri |
|
2025_12_Wu_Thesis.pdf
non accessibile
Descrizione: Thesis
Dimensione
20.52 MB
Formato
Adobe PDF
|
20.52 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/246755