Reusable space launch vehicles are a key enabler for reducing the cost of access to space and represent a central objective of current European initiatives such as the SALTO mission. These technologies impose stringent requirements on onboard guidance algorithms during the powered descent and landing phase, which is typically formulated as a fuel-optimal nonlinear Optimal Control Problem (OCP). In recent years, Successive Convexification (SCvx) has emerged as a state-of-the-art approach for solving Powered Descent Guidance (PDG) problems in real time, commonly relying on Second-Order Cone Programming (SOCP) sub-problems, which, however, often incur a significant computational burden. This thesis investigates an alternative SCvx formulation for PDG based exclusively on con- vex Quadratic Programming (QP) sub-problems. Starting from a five Degrees-of-Freedom (DoF) powered landing OCP with aerodynamic effects and angle-of-attack constraints, the original nonconvex problem is systematically convexified using linearizations, virtual controls, and trust regions, and then reformulated into canonical dense and optimal-control QP structures. The proposed framework is implemented and tested using multiple off-the-shelf QP solvers, and is compared against a SOCP baseline. The results and performance are assessed on the basis of a realistic reusable launcher case study from the RETALT mission. The analysis considers convergence behavior, computational cost, solution accuracy, robustness to initialization and model uncertainties, and sensitivity to solver parameters, including Monte Carlo simulations. Overall, results show that QP-based SCvx formulations can achieve comparable solution quality and robustness to SOCP-based methods, while offering substantial reductions in computational time and improved suitability for real-time onboard applications.
Il riutilizzo di lanciatori spaziali rappresenta un elemento chiave per la riduzione dei costi di accesso allo spazio e costituiscono un obiettivo centrale delle iniziative europee, come la missione SALTO. Tali tecnologie impongono requisiti stringenti agli algoritmi di guida di bordo durante la fase di discesa e atterraggio propulso, tipicamente formulata come un problema di controllo ottimo (OCP) non lineare a minimo consumo di propellente. Negli ultimi anni, la convessificazione successiva (SCvx) si è affermata come approccio di riferimento per la risoluzione in tempo reale dei problemi di guida di discesa propulsa (PDG), facendo comunemente ricorso a sottoproblemi di programmazione a coni di secondo ordine (SOCP), che tuttavia comportano spesso un significativo carico computazionale. Questa tesi indaga una formulazione alternativa della SCvx per la PDG basata esclusivamente su sottoproblemi di programmazione quadratica (QP) convessa. A partire da un OCP di atterraggio propulso a cinque gradi di libertà (DoF), comprensivo di effetti aerodinamici e vincoli sull’angolo d’attacco, il problema non convesso originale viene sistematicamente convessificato mediante linearizzazioni, controlli virtuali e regioni di fiducia, per poi essere riformulato in strutture QP canoniche di tipo denso e in forma di controllo ottimo. La struttura proposta viene implementata e testata utilizzando diversi risolutori QP commerciali e confrontata con una formulazione di riferimento basata su SOCP. I risultati vengono valutati sulla base di un caso di studio realistico relativo a un lanciatore spaziale riutilizzabile nell’ambito della missione RETALT. L’analisi considera il comportamento di convergenza, il costo computazionale, l’accuratezza della soluzione e la robustezza rispetto all’inizializzazione e alle incertezze di modello, includendo simulazioni Monte Carlo. Nel complesso, i risultati mostrano che le formulazioni SCvx basate su QP raggiungono prestazioni comparabili a quelle dei metodi basati su SOCP, offrendo al contempo una significativa riduzione dei tempi di calcolo e una migliore idoneità all’implementazione in tempo reale su sistemi di bordo.
Successive convexification with quadratic programming for powered landing guidance of reusable launchers
Portantiolo, Matteo
2024/2025
Abstract
Reusable space launch vehicles are a key enabler for reducing the cost of access to space and represent a central objective of current European initiatives such as the SALTO mission. These technologies impose stringent requirements on onboard guidance algorithms during the powered descent and landing phase, which is typically formulated as a fuel-optimal nonlinear Optimal Control Problem (OCP). In recent years, Successive Convexification (SCvx) has emerged as a state-of-the-art approach for solving Powered Descent Guidance (PDG) problems in real time, commonly relying on Second-Order Cone Programming (SOCP) sub-problems, which, however, often incur a significant computational burden. This thesis investigates an alternative SCvx formulation for PDG based exclusively on con- vex Quadratic Programming (QP) sub-problems. Starting from a five Degrees-of-Freedom (DoF) powered landing OCP with aerodynamic effects and angle-of-attack constraints, the original nonconvex problem is systematically convexified using linearizations, virtual controls, and trust regions, and then reformulated into canonical dense and optimal-control QP structures. The proposed framework is implemented and tested using multiple off-the-shelf QP solvers, and is compared against a SOCP baseline. The results and performance are assessed on the basis of a realistic reusable launcher case study from the RETALT mission. The analysis considers convergence behavior, computational cost, solution accuracy, robustness to initialization and model uncertainties, and sensitivity to solver parameters, including Monte Carlo simulations. Overall, results show that QP-based SCvx formulations can achieve comparable solution quality and robustness to SOCP-based methods, while offering substantial reductions in computational time and improved suitability for real-time onboard applications.| File | Dimensione | Formato | |
|---|---|---|---|
|
2026_03_Portantiolo_Tesi.pdf
accessibile in internet per tutti
Descrizione: testo tesi
Dimensione
9.52 MB
Formato
Adobe PDF
|
9.52 MB | Adobe PDF | Visualizza/Apri |
|
2026_03_Portantiolo_Executive Summary.pdf
accessibile in internet per tutti
Descrizione: executive summary
Dimensione
1.11 MB
Formato
Adobe PDF
|
1.11 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/10589/253794