This thesis is about low-thrust Space Trajectories Optimisation, which consists of finding the thrust profile that minimises the fuel required by a spacecraft during its transfer. Many methods are used in literature to solve such a problem: indirect methods exploit the Euler - Lagrange theorem to formulate the necessary conditions for optimality starting from the optimal control problem; they are able to achieve great results in terms of optimality, however they require a very good initial guess to reach convergence. Direct methods transform instead the optimal control problem into a set of discretized constraints; they require a less accurate initial guess but they are able to reach less optimal solutions as well. The EXTREMA project by Professor Francesco Topputo aims at disrupting the current paradigm of how space trajectories are optimised: nowadays, engineers must re-design a S/C trajectory every time it deviates from the nominal conditions and this is a costly and time-consuming process; EXTREMA aims at developing an autonomous navigation system for interplanetary CubeSats missions, capable of self-re designing the S/C trajectory when off-nominal conditions are encountered by means of a so-called closed-loop guidance algorithm. However, when it comes to solving the problem of finding the optimal space trajectory directly on-board a CubeSat during its interplanetary trajectory, state-of-the-art direct and indirect methods are not suitable; in fact, they are not able to provide the following three fundamental characteristics at the same time: robustness, optimality and sustainability. Convex Optimisation, a relatively new approach to solve optimal control problems, may be the most appropriate technique for closed-loop guidance: it is sustainable since it solves problems by means of polynomial-time algorithms and it has acceptable levels of robustness and optimality. Among the three aforementioned characteristics, sustainability and robustness are of paramount importance: the on-board computer must find, with the available resources, a feasible trajectory to follow at any time instant. In this work, we focused on increasing these two properties of the Sequential Convex Programming algorithm, which is based on Convex Optimisation: we merged it together with what is commonly referred to as Homotopic Approach. The Homotopic Approach applied to indirect methods basically consists of first solving smoother problems and finally providing the fuel-optimal one, which is difficult to solve due to its discontinuous solution, with a very accurate initial guess obtained by solving the aforementioned smoother problems. In this work we investigate instead to what extent a combination of Sequential Convex Programming and Homotopic Approach is able to improve the results obtained through state-of-the-art methods. We answer to this research question by means of two objectives: 1. build a Gauss - Lobatto-based Sequential Convex Programming algorithm that includes the Homotopic Approach technique; 2. simulate spacecraft off-nominal conditions during its journey and successfully re-design its optimal trajectory. We show here that the two objectives have been satisfied and that our approach is actually potentially able to overcome state-of-the-art results: in more than 95% of the cases, it shown better or equal convergence properties than the standard approach did. Finally, we provide some insights on future research based on our work.
Questa tesi tratta il problema dell’ottimizzazione delle traiettorie spaziali, che consiste nel definire il profilo di spinta capace di minimizzare il combustibile richiesto da un veicolo spaziale durante il suo trasferimento dalla Terra a un corpo celeste. In letteratura, sono diversi i metodi numerici utilizzati per risolvere il problema: i metodi indiretti sfruttano il teorema di Eulero - Lagrange per formulare le condizioni necessarie di ottimalità a partire dalla formulazione del problema di controllo ottimo; questi metodi sono capaci di raggiungere alti livelli di ottimalità della soluzione, ma necessitano di una ipotesi iniziale di traiettoria molto accurata. I metodi diretti, invece, trasformano il problema di controllo ottimo in un insieme di vincoli discreti; essi invece richiedono un’ ipotesi iniziale di traiettoria meno precisa, ma raggiungono livelli di ottimalità più bassi rispetto ai metodi indiretti. Il progetto EXTREMA, del Professor Francesco Topputo, ha come obiettivo rivoluzionare il paradigma attuale di come le traiettorie spaziali vengono ottimizzate: allo stato attuale, gli ingegneri hanno il compito di calcolare nuovamente la traiettoria di un veicolo spaziale ogniqualvolta essa si discosti dalle sue condizioni nominali, e questo è un compito economicamente costoso che richiede una discreta quantità di tempo per essere svolto. EXTREMA vuole sviluppare un sistema di navigazione autonoma per CubeSats che sia capace di ricalclolare autonomamente la propria traiettoria quando condizioni non nominali si verifichino. Tuttavia, i classici metodi diretti e indiretti non sono adatti a questo tipo di applicazioni, perché non sono in grado di garantire, allo stesso tempo, le seguenti tre caratteristiche fondamentali: robustezza, ottimalità e sostenibilità degli algoritmi. L’Ottimizzazione Convessa potrebbe invece rappresentare l’approccio più adatto all’ottimizzazione, in tempo reale e a bordo del veicolo spaziale, della sua traiettoria: essa garantisce la caratteristica di sostenibilità grazie al fatto che i problemi possono essere risolti per mezzo di questa tecnica attraverso algoritmi che richiedono tempi polinomiali; inoltre, essa possiede discreti livelli di ottimalià e robustezza. Tra le tre caratteristiche fondamentali già menzionate, robustezza e sostenibilità rappresentano quelle di maggiore importanza: il computer di bordo deve essere in grado di trovare, in ogni istante di tempo e con le risorse disponibili, una traiettoria da seguire. Questa tesi si concentra sulle strategie atte a incrementare robustezza e sostenibilità del cosiddetto algoritmo Sequential Convex Programming, che si basa sull’Ottimizzazione Convessa: ad esso abbiamo applicato un metodo chiamato, in Inglese, Homotopic Approach; esso consiste nel risolvere problemi più semplici rispetto a quello originale, per poi fornire a quest’ultimo la soluzione ottenuta dai problemi più semplici come ipotesi iniziale della traiettoria. In particolare, in questo lavoro investighiamo fino a che punto una combinazione di Sequential Convex Programming e Homotopic Approach sia in grado di migliorare i risultati ottenuti con metodi classici. Risponderemo a questa domanda di ricerca attraverso i seguenti due obiettivi: 1. lo sviluppo di un algoritmo di Sequential Convex Programming basato sul metodo di Gauss - Lobatto che includa anche l’Homotopic Approach; 2. la simulazione di ricalcolo della traiettoria di un veicolo spaziale dopo che esso incontra condizioni non nominali. Dimostreremo che il nostro approccio è potenzialmente in grado di migliorare i risultati ottenuti con metodi classici, in quanto ha mostrato, in più del 95% dei casi, proprietà di convergenza migliori o uguali rispetto agli approcci standard.
Robust design of low-thrust minimum-fuel space trajectories by combination of sequential convex programming and homotopic approach
Morelli, Andrea Carlo
2019/2020
Abstract
This thesis is about low-thrust Space Trajectories Optimisation, which consists of finding the thrust profile that minimises the fuel required by a spacecraft during its transfer. Many methods are used in literature to solve such a problem: indirect methods exploit the Euler - Lagrange theorem to formulate the necessary conditions for optimality starting from the optimal control problem; they are able to achieve great results in terms of optimality, however they require a very good initial guess to reach convergence. Direct methods transform instead the optimal control problem into a set of discretized constraints; they require a less accurate initial guess but they are able to reach less optimal solutions as well. The EXTREMA project by Professor Francesco Topputo aims at disrupting the current paradigm of how space trajectories are optimised: nowadays, engineers must re-design a S/C trajectory every time it deviates from the nominal conditions and this is a costly and time-consuming process; EXTREMA aims at developing an autonomous navigation system for interplanetary CubeSats missions, capable of self-re designing the S/C trajectory when off-nominal conditions are encountered by means of a so-called closed-loop guidance algorithm. However, when it comes to solving the problem of finding the optimal space trajectory directly on-board a CubeSat during its interplanetary trajectory, state-of-the-art direct and indirect methods are not suitable; in fact, they are not able to provide the following three fundamental characteristics at the same time: robustness, optimality and sustainability. Convex Optimisation, a relatively new approach to solve optimal control problems, may be the most appropriate technique for closed-loop guidance: it is sustainable since it solves problems by means of polynomial-time algorithms and it has acceptable levels of robustness and optimality. Among the three aforementioned characteristics, sustainability and robustness are of paramount importance: the on-board computer must find, with the available resources, a feasible trajectory to follow at any time instant. In this work, we focused on increasing these two properties of the Sequential Convex Programming algorithm, which is based on Convex Optimisation: we merged it together with what is commonly referred to as Homotopic Approach. The Homotopic Approach applied to indirect methods basically consists of first solving smoother problems and finally providing the fuel-optimal one, which is difficult to solve due to its discontinuous solution, with a very accurate initial guess obtained by solving the aforementioned smoother problems. In this work we investigate instead to what extent a combination of Sequential Convex Programming and Homotopic Approach is able to improve the results obtained through state-of-the-art methods. We answer to this research question by means of two objectives: 1. build a Gauss - Lobatto-based Sequential Convex Programming algorithm that includes the Homotopic Approach technique; 2. simulate spacecraft off-nominal conditions during its journey and successfully re-design its optimal trajectory. We show here that the two objectives have been satisfied and that our approach is actually potentially able to overcome state-of-the-art results: in more than 95% of the cases, it shown better or equal convergence properties than the standard approach did. Finally, we provide some insights on future research based on our work.File | Dimensione | Formato | |
---|---|---|---|
Thesis_Morelli_AndreaCarlo.pdf
accessibile in internet per tutti
Descrizione: Thesis Document
Dimensione
3.37 MB
Formato
Adobe PDF
|
3.37 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/10589/174899