The growth of satellites in space mirrors their crucial role in modern society, driving navigation, telecommunication, and Earth observation services. As new satellite constellations emerge, missions such as active debris removal, multi-injection, and in-orbit servicing gain prominence. A recurring theme in their mission analysis is the multi-target aspect, allowing a single spacecraft to visit multiple objects in one mission, enabled by using high-impulse, low-thrust propulsion systems. However, the trajectory design presents challenges, including identifying optimal target combinations within a vast search space and the computational demands of designing low-thrust trajectories. In this thesis, a trajectory design methodology is proposed, that uses artificial neural networks to estimate propellant mass consumption and time of flight related to orbit-to-orbit transfers. Each transfer is composed of three parts: the first and the third are low-thrust legs, calculated using a three-dimensional shape-based algorithm. This algorithm is based on non-linear interpolation of consecutive orbits, allowing it to function even in scenarios involving hundreds of revolutions typical of Earth-centered missions. These two are connected by a coasting arc, which exploits the J_2 perturbation to change the Right Ascension of Ascending Node (RAAN). Particle Swarm Optimisation (PSO) is used to optimise the altitude and inclination of the intermediate drifting orbit. Initially, a pool of optimal transfers between randomly-generated Earth orbits is created. These are used to train an artificial neural network through supervised learning. The structure and key parameters of the network are optimised using Optuna. The proposed methodology, which combines the trained artificial neural network with a breadth-first tree search to address the combinatorial aspect of the problem, is evaluated considering an in-orbit servicing mission involving a search space of 200 randomly-generated objects. The results are assessed in terms of accuracy, computational burden, and time, and compared against the method combining the shape-based algorithm and the PSO used for database generation. While demonstrating clear computational advantages, the trained neural network exhibits higher error rates compared to those typically reported in the literature, indicating areas for further research and refinement.
La crescita dei satelliti nello spazio riflette il loro ruolo cruciale nella società moderna, supportando servizi di navigazione, telecomunicazione e osservazione della Terra. Con l'emergere di nuove costellazioni satellitari, missioni come la rimozione attiva dei detriti, la multi-iniezione e il servicing in orbita guadagnano importanza. Un tema ricorrente dalla prospettiva dell'analisi di missione è l'aspetto multi-target, che consente a un singolo veicolo spaziale di visitare più oggetti in un'unica missione, reso possibile dall'uso di sistemi di propulsione ad alto impulso e bassa spinta. Tuttavia, la progettazione della traiettoria presenta sfide, tra cui l'identificazione di combinazioni ottimali di obiettivi all'interno di un vasto spazio di ricerca e gli oneri computazionali per la progettazione di traiettorie a bassa spinta. In questa tesi, viene proposta una metodologia di progettazione della traiettoria che utilizza reti neurali artificiali per stimare il consumo di massa di propellente e il tempo di volo relativi ai trasferimenti orbita-orbita. Ogni trasferimento è composto da tre parti: la prima e la terza sono gambe a bassa spinta, calcolate utilizzando un algoritmo shape-based tridimensionale. Questo algoritmo si basa sull'interpolazione non lineare di orbite consecutive, consentendo il suo funzionamento anche in scenari che coinvolgono centinaia di rivoluzioni tipiche delle missioni con la Terra come corpo centrale. Questi due sono collegati da un arco di deriva, che sfrutta la perturbazione J_2 per cambiare l'ascensione retta del nodo ascendente (RAAN). La Particle Swarm Optimisation (PSO) viene utilizzata per ottimizzare l'altitudine e l'inclinazione dell'orbita intermedia di deriva. Inizialmente, viene creato un pool di trasferimenti ottimali tra orbite terrestri generate casualmente. Questi vengono utilizzati per addestrare una rete neurale artificiale tramite apprendimento supervisionato. La struttura e i parametri chiave della rete vengono ottimizzati utilizzando Optuna. La metodologia proposta, che combina la rete neurale artificiale con una ricerca ad albero in ampiezza per affrontare l'aspetto combinatorio del problema, viene valutata considerando una missione di servicing in orbita che coinvolge uno spazio di ricerca di 200 oggetti generati casualmente. I risultati sono valutati in termini di accuratezza, carico computazionale e tempo, e confrontati con il metodo che combina l'algoritmo shape-based e il PSO utilizzati per la generazione del database. Nonostante dimostri evidenti vantaggi computazionali, la rete neurale addestrata presenta tassi di errore più elevati rispetto a quelli tipicamente riportati in letteratura, indicando aree per ulteriori ricerche e affinamenti.
Earth multi-target trajectory design with artificial neural networks
Barbieri, Anna
2022/2023
Abstract
The growth of satellites in space mirrors their crucial role in modern society, driving navigation, telecommunication, and Earth observation services. As new satellite constellations emerge, missions such as active debris removal, multi-injection, and in-orbit servicing gain prominence. A recurring theme in their mission analysis is the multi-target aspect, allowing a single spacecraft to visit multiple objects in one mission, enabled by using high-impulse, low-thrust propulsion systems. However, the trajectory design presents challenges, including identifying optimal target combinations within a vast search space and the computational demands of designing low-thrust trajectories. In this thesis, a trajectory design methodology is proposed, that uses artificial neural networks to estimate propellant mass consumption and time of flight related to orbit-to-orbit transfers. Each transfer is composed of three parts: the first and the third are low-thrust legs, calculated using a three-dimensional shape-based algorithm. This algorithm is based on non-linear interpolation of consecutive orbits, allowing it to function even in scenarios involving hundreds of revolutions typical of Earth-centered missions. These two are connected by a coasting arc, which exploits the J_2 perturbation to change the Right Ascension of Ascending Node (RAAN). Particle Swarm Optimisation (PSO) is used to optimise the altitude and inclination of the intermediate drifting orbit. Initially, a pool of optimal transfers between randomly-generated Earth orbits is created. These are used to train an artificial neural network through supervised learning. The structure and key parameters of the network are optimised using Optuna. The proposed methodology, which combines the trained artificial neural network with a breadth-first tree search to address the combinatorial aspect of the problem, is evaluated considering an in-orbit servicing mission involving a search space of 200 randomly-generated objects. The results are assessed in terms of accuracy, computational burden, and time, and compared against the method combining the shape-based algorithm and the PSO used for database generation. While demonstrating clear computational advantages, the trained neural network exhibits higher error rates compared to those typically reported in the literature, indicating areas for further research and refinement.File | Dimensione | Formato | |
---|---|---|---|
2024_04_Barbieri_Executive_Summary_02.pdf
accessibile in internet per tutti
Dimensione
705.65 kB
Formato
Adobe PDF
|
705.65 kB | Adobe PDF | Visualizza/Apri |
2024_04_Barbieri_Tesi_01.pdf
accessibile in internet per tutti
Dimensione
5.81 MB
Formato
Adobe PDF
|
5.81 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/10589/219818