This thesis develops and compares control strategies to enhance Lateral Vehicle Motion Control (LVMC) performance in an SUV equipped with front and rear Steer-by-Wire (SBW), Brake-by-Wire (BBW), and semi-active dampers. The control variables are primarily managed by a linear Model Predictive Controller (MPC), formulated via quadratic programming, with additional modules incorporated when necessary. Vehicle simulations are conducted using VI-CarRealTime (VI-CRT) software. The first established strategies serve as baselines and consist of a heuristic damping module and a parallel MPC that controls either the front SBW or both the front and rear SBW. Thus, both a linear and a non-linear Single-Track Model (STM) are developed, identified, and validated, with the linear STM ultimately embedded in the baseline MPC. The MPC requires reference signals to follow, generated via a dedicated subsystem fed with the predicted driver’s request at the steering wheel. Two alternative prediction strategies are tested, based on constant and parabolic extrapolation. Then, two alternative architectures are proposed to improve lateral control. The first integrates the BBW system into the baseline MPC, complemented by a brake pressure allocator. The second introduces a centralized MPC, coordinating all control variables, thus eliminating the heuristic damping module. This centralized approach leverages the coupling of lateral and vertical dynamics via roll motion, leading to the development of the Roll Model (RM), incorporating the roll angle and its derivative as state variables. A sensitivity analysis determines the optimal prediction horizon for each configuration. The final comparative analysis highlights the benefits of including the BBW system rather than employing a centralized MPC with RM. Indeed, the best architecture, including an MPC controlling the BBW and both front and rear SBW, outperforms the baseline, the MPC governs only the front SBW. This results in reductions of 11.5% and 25.9% in yaw rate average and peak tracking errors, respectively, and 50.7% and 57.5% in side-slip angle average and peak tracking errors.
Questa tesi si concentra sullo sviluppo e confronto di strategie di controllo per migliorare le prestazioni del Lateral Vehicle Motion Control (LVMC) in un SUV dotato di Steer-by-Wire (SBW) anteriore e posteriore, Brake-by-Wire (BBW) e smorzatori semi-attivi. Le variabili di controllo sono gestite principalmente da un Model Predictive Controller (MPC) lineare, formulato tramite Quadratic Programming, e moduli aggiuntivi integrati quando necessario. Le simulazioni del veicolo sono condotte tramite il software VI-CarRealTime (VI-CRT). La prima strategia di controllo implementata funge da riferimento e include un modulo euristico per la gestione dei coefficienti di smorzamento e un MPC che fornisce gli angoli di sterzo anteriore e posteriore. A tal fine sono stati sviluppati e validati due Single-Track Model (STM), non lineare e lineare, con preferenza per quest’ultimo poiché più semplice ma ugualmente efficace. L’MPC richiede valori di riferimento, generati tramite un modulo che riceve in ingresso la previsione della richiesta del conducente al volante, calcolata con tecniche di estrapolazione costante e parabolica. Sono state poi realizzate due architetture alternative per migliorare il controllo della dinamica laterale. La prima integra il sistema BBW nell’MPC di riferimento, affiancato da un allocatore di pressione frenante. La seconda impiega un MPC centralizzato per coordinare tutte le variabili, rendendo il modulo euristico di smorzamento non necessario. Questo approccio centralizzato sfrutta l’accoppiamento delle dinamiche laterali e verticali tramite il rollio della cassa e degli assali, portando allo sviluppo del Roll Model (RM). Per ciascuna configurazione è stata successivamente condotta un’analisi di sensitività per determinare la lunghezza ottimale dell’orizzonte di previsione di ciascun MPC. Infine, l’analisi comparativa tra queste architetture di controllo dimostra che l’integrazione del sistema BBW nella configurazione di controllo di riferimento offre vantaggi significativi rispetto agli MPC che combinano le dinamiche laterali e verticali tramite RM. Infatti la migliore architettura di controllo, che include l’attuazione dei freni e dello sterzo anteriore e posteriore, comporta una riduzione, rispetto alla configurazione di riferimento che utilizza soltanto lo sterzo anteriore, dell’11.5% e del 25.9% rispettivamente per l’errore medio e di picco dell’inseguimento di yaw rate, e del 50.7% e 57.5% per l’errore medio e di picco del side-slip angle.
Design and comparative analysis of multi-actuator model predictive control systems for lateral vehicle motion control
Bavaresco, Giuseppe;Canzii, Gianluca
2023/2024
Abstract
This thesis develops and compares control strategies to enhance Lateral Vehicle Motion Control (LVMC) performance in an SUV equipped with front and rear Steer-by-Wire (SBW), Brake-by-Wire (BBW), and semi-active dampers. The control variables are primarily managed by a linear Model Predictive Controller (MPC), formulated via quadratic programming, with additional modules incorporated when necessary. Vehicle simulations are conducted using VI-CarRealTime (VI-CRT) software. The first established strategies serve as baselines and consist of a heuristic damping module and a parallel MPC that controls either the front SBW or both the front and rear SBW. Thus, both a linear and a non-linear Single-Track Model (STM) are developed, identified, and validated, with the linear STM ultimately embedded in the baseline MPC. The MPC requires reference signals to follow, generated via a dedicated subsystem fed with the predicted driver’s request at the steering wheel. Two alternative prediction strategies are tested, based on constant and parabolic extrapolation. Then, two alternative architectures are proposed to improve lateral control. The first integrates the BBW system into the baseline MPC, complemented by a brake pressure allocator. The second introduces a centralized MPC, coordinating all control variables, thus eliminating the heuristic damping module. This centralized approach leverages the coupling of lateral and vertical dynamics via roll motion, leading to the development of the Roll Model (RM), incorporating the roll angle and its derivative as state variables. A sensitivity analysis determines the optimal prediction horizon for each configuration. The final comparative analysis highlights the benefits of including the BBW system rather than employing a centralized MPC with RM. Indeed, the best architecture, including an MPC controlling the BBW and both front and rear SBW, outperforms the baseline, the MPC governs only the front SBW. This results in reductions of 11.5% and 25.9% in yaw rate average and peak tracking errors, respectively, and 50.7% and 57.5% in side-slip angle average and peak tracking errors.File | Dimensione | Formato | |
---|---|---|---|
2024_12_Bavaresco_Canzii_Executive_Summary.pdf
non accessibile
Descrizione: Executive Summary
Dimensione
1.35 MB
Formato
Adobe PDF
|
1.35 MB | Adobe PDF | Visualizza/Apri |
2024_12_Bavaresco_Canzii_Tesi.pdf
non accessibile
Descrizione: Tesi
Dimensione
26.26 MB
Formato
Adobe PDF
|
26.26 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/10589/230508