The exponential growth of the Resident Space Object (RSO) population and the increasing congestion of the orbital environment pose significant challenges for Space Surveillance and Tracking (SST) systems. Essential to the safety of space operations is the maintenance of accurate catalogues; however, this task is frequently disrupted by unmodelled satellite manoeuvres, which render standard ballistic propagation models obsolete. Although the problem has been addressed in the literature, within the routine tracking domain, existing solutions often rely on computationally intensive techniques, such as filter banks, that limit their scalability in operational scenarios. To address this need for operational efficiency, this thesis presents a novel methodology for the detection and estimation of manoeuvres, specifically designed to be integrated into a standard Weighted Least Squares (WLSQ) batch Orbit Determination (OD) process. Unlike the aforementioned computationally heavy approaches, the proposed algorithm utilises the improvement of the predicted Weighted Root Mean Square (WRMS) of the residuals when assuming a manoeuvre as a trigger for detection. Upon identifying a potential anomaly, the method executes a parametric sweep over the fitting interval to estimate the manoeuvre epoch and magnitude via a differential correction process, without relying on an accurate pre-manoeuvre orbit estimate. A key innovation of this research lies in its computational efficiency. By reusing the partial derivative matrices and propagation data generated during the nominal OD iterations, the algorithm achieves manoeuvre characterisation with negligible additional computational cost. This feature makes it uniquely suitable for high-volume operational environments where processing time is a critical constraint. The performance of the algorithm was validated through a comprehensive Monte Carlo campaign involving synthetic Low Earth Orbit (LEO) scenarios. Results demonstrate that the method achieves high classification metrics (F1 score) while maintaining a low false positive rate. The findings indicate that the proposed strategy effectively bridges the gap between high-fidelity estimation and operational efficiency, enabling the automated generation of manoeuvre trend patterns to support future track-to-orbit association tasks.
La crescita esponenziale della popolazione di oggetti spaziali (RSO) e la crescente congestione dell'ambiente orbitale pongono sfide significative nell'ambito del Space Surveillance and Tracking (SST). Essenziale per la sicurezza delle operazioni spaziali è il mantenimento di cataloghi accurati; tuttavia, questo compito è frequentemente ostacolato da manovre satellitari non modellate, che rendono obsoleti i modelli di propagazione balistica standard. Sebbene il problema sia stato affrontato in letteratura, nell'ambito del tracking ordinario, le soluzioni esistenti si affidano spesso a tecniche computazionalmente onerose, come i banchi di filtri, che ne limitano la scalabilità in scenari operativi. Per rispondere a questa esigenza di efficienza operativa, questa tesi presenta un'innovativa metodologia per il rilevamento e la stima delle manovre, progettata specificamente per essere integrata in un processo standard di determinazione orbitale (OD) batch Least Squares (WLSQ). A differenza dei suddetti approcci computazionalmente pesanti, l'algoritmo proposto utilizza, come innesco per il rilevamento, il miglioramento del valore previsto del Weighted Root Mean Square (WRMS) dei residui quando si ipotizza una manovra. Una volta identificata una potenziale anomalia, il metodo esegue una scansione parametrica sull'intervallo di fitting per stimare l'epoca e la grandezza della manovra attraverso un processo di correzione differenziale, senza fare affidamento su una stima accurata dell'orbita pre-manovra. Riutilizzando le matrici delle derivate parziali e i dati di propagazione generati durante le iterazioni nominali di OD, l'algoritmo caratterizza la manovra con un costo computazionale aggiuntivo trascurabile. Le prestazioni del metodo sono state validate attraverso una completa analisi Monte Carlo su scenari sintetici in orbita terrestre bassa (LEO). I risultati dimostrano che il metodo raggiunge elevate metriche di classificazione (F1 score) mantenendo al contempo un basso tasso di falsi positivi. Le conclusioni indicano che la strategia proposta colma efficacemente il divario tra stima ad alta fedeltà ed efficienza operativa, consentendo la generazione automatizzata di storici delle manovre a supporto di futuri compiti di associazione track-to-orbit.
Efficient satellite manoeuvre detection in batch orbit determination processes
Alonso Vélez, Carlos
2025/2026
Abstract
The exponential growth of the Resident Space Object (RSO) population and the increasing congestion of the orbital environment pose significant challenges for Space Surveillance and Tracking (SST) systems. Essential to the safety of space operations is the maintenance of accurate catalogues; however, this task is frequently disrupted by unmodelled satellite manoeuvres, which render standard ballistic propagation models obsolete. Although the problem has been addressed in the literature, within the routine tracking domain, existing solutions often rely on computationally intensive techniques, such as filter banks, that limit their scalability in operational scenarios. To address this need for operational efficiency, this thesis presents a novel methodology for the detection and estimation of manoeuvres, specifically designed to be integrated into a standard Weighted Least Squares (WLSQ) batch Orbit Determination (OD) process. Unlike the aforementioned computationally heavy approaches, the proposed algorithm utilises the improvement of the predicted Weighted Root Mean Square (WRMS) of the residuals when assuming a manoeuvre as a trigger for detection. Upon identifying a potential anomaly, the method executes a parametric sweep over the fitting interval to estimate the manoeuvre epoch and magnitude via a differential correction process, without relying on an accurate pre-manoeuvre orbit estimate. A key innovation of this research lies in its computational efficiency. By reusing the partial derivative matrices and propagation data generated during the nominal OD iterations, the algorithm achieves manoeuvre characterisation with negligible additional computational cost. This feature makes it uniquely suitable for high-volume operational environments where processing time is a critical constraint. The performance of the algorithm was validated through a comprehensive Monte Carlo campaign involving synthetic Low Earth Orbit (LEO) scenarios. Results demonstrate that the method achieves high classification metrics (F1 score) while maintaining a low false positive rate. The findings indicate that the proposed strategy effectively bridges the gap between high-fidelity estimation and operational efficiency, enabling the automated generation of manoeuvre trend patterns to support future track-to-orbit association tasks.| File | Dimensione | Formato | |
|---|---|---|---|
|
2026_03_Alonso_Thesis_01.pdf
accessibile in internet per tutti
Descrizione: Text of the thesis
Dimensione
3.96 MB
Formato
Adobe PDF
|
3.96 MB | Adobe PDF | Visualizza/Apri |
|
2026_03_Alonso_Executive Summary_02.pdf
accessibile in internet per tutti
Descrizione: Executive summary
Dimensione
2.17 MB
Formato
Adobe PDF
|
2.17 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/10589/252763