Recent technological and engineering advances are enabling higher performance in many domains, posing new challenges for control design. In practice, the exact dynamics of the system to be controlled are often unknown, and using highly detailed models in control loops typically results in prohibitive computational costs, especially for real-time or safety-critical applications. This thesis investigates a high-performance control strategy for a sport vehicle system using Koopman operator-based modeling and control. This approach has attracted growing interest due to its ability to capture rich nonlinear system behavior by embedding it into a higher-dimensional linear system, enabling efficient computation and simplified control design. We present the implementation of a Koopman-based Model Predictive Control strategy in a quasi-realistic vehicular context, discussing the practical challenges encountered in the process. The results highlight how this method combines the accuracy of nonlinear MPC with the computational efficiency of linear time-invariant MPC, suggesting that Koopman-based control may offer a promising trade-off between performance and feasibility in embedded, real-time applications.
I recenti sviluppi tecnologici e ingegneristici stanno permettendo di raggiungere prestazioni sempre più elevate in numerosi ambiti, ponendo nuove sfide nella progettazione dei sistemi di controllo. Nel pratico, le dinamiche esatte del sistema da controllare non sono spesso note, e l’utilizzo di modelli molto dettagliati nei sistemi di controllo comporta solitamente tempi computazionali proibitivi, soprattutto nei contesti safety-critical o che richiedono risposte in tempo reale. Questa tesi esplora una strategia di controllo ad alte prestazioni applicata a un veicolo sportivo, basata sulla modellazione e sul controllo tramite operatore di Koopman. Questo approccio ha suscitato un crescente interesse grazie alla sua capacità di catturare il comportamento non lineare del sistema tramite una proiezione su uno spazio lineare a dimensione superiore, che consente una progettazione del controllo più semplice ed efficiente dal punto di vista computazionale. Presentiamo l’implementazione di un controllo predittivo (MPC) basato su Koopman in un contesto veicolare quasi-realistico, discutendo le sfide pratiche emergono durante il processo. I risultati evidenziano come il metodo combini il vantaggio di accuratezza dell'MPC nonlineare con l’efficienza computazionale dell'MPC lineare tempo-invariante.
Koopman-based MPC for high performance vehicles
De Pasquale, Zanubia
2024/2025
Abstract
Recent technological and engineering advances are enabling higher performance in many domains, posing new challenges for control design. In practice, the exact dynamics of the system to be controlled are often unknown, and using highly detailed models in control loops typically results in prohibitive computational costs, especially for real-time or safety-critical applications. This thesis investigates a high-performance control strategy for a sport vehicle system using Koopman operator-based modeling and control. This approach has attracted growing interest due to its ability to capture rich nonlinear system behavior by embedding it into a higher-dimensional linear system, enabling efficient computation and simplified control design. We present the implementation of a Koopman-based Model Predictive Control strategy in a quasi-realistic vehicular context, discussing the practical challenges encountered in the process. The results highlight how this method combines the accuracy of nonlinear MPC with the computational efficiency of linear time-invariant MPC, suggesting that Koopman-based control may offer a promising trade-off between performance and feasibility in embedded, real-time applications.| File | Dimensione | Formato | |
|---|---|---|---|
|
Thesis - Koopman-based MPC for High Performance Vehicles.pdf
non accessibile
Descrizione: Thesis
Dimensione
10.39 MB
Formato
Adobe PDF
|
10.39 MB | Adobe PDF | Visualizza/Apri |
|
Executive Summary - Koopman-based MPC for High Performance Vehicles.pdf
non accessibile
Descrizione: Executive Summary
Dimensione
1.89 MB
Formato
Adobe PDF
|
1.89 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/10589/240388