This thesis investigates the problem of Autonomous Air-to-Air Landing (AAAL), in which a small unmanned aerial vehicle (UAV) autonomously lands on a larger carrier drone in motion. This complex task integrates vision-based navigation, state estimation, and control under dynamic conditions. Building upon previous research conducted at the Politecnico di Milano, three main contributions are presented. First, an Adaptive Input and State Estimation (AISE) filter is implemented as an alternative to the conventional Kalman Filter (KF). Unlike the KF, the AISE filter simultaneously estimates both velocity and acceleration states. Second, two harmonic disturbance compensators are developed, a Standard Harmonic Compensator (SHC) and an Adaptive Harmonic Compensator (AHC), to improve robustness against oscillatory disturbances and ensure accurate tracking of harmonic trajectories. Third, a simulation environment is designed in Simulink using the Unreal Engine interface, enabling realistic visualization of 3D landing scenarios, including camera-based ArUco marker detection and dynamic interactions between UAVs. The proposed estimation and control architectures are extensively evaluated through simulations involving various carrier trajectories, from rectilinear to complex circular and figure-eight paths. Results indicate that the Kalman Filter outperforms the AISE in velocity estimation, resulting in faster and more stable landings. Among the compensators, the SHC exhibited the highest robustness, achieving successful landings even at increased carrier angular velocities. The developed simulation framework proved to be an effective and versatile platform for testing vision-based estimation and control strategies in AAAL operations. Finally, in-flight experiments validated the superior performance of the proposed compensators compared to previous adaptive controllers, confirming the robustness of the SHC and the adaptability of the AHC. Overall, this work advances the development of autonomous landing technologies for UAVs by introducing enhanced estimation and disturbance rejection methods and by providing a validated simulation environment that bridges the gap between theory and experimental implementation.
Questa tesi indaga il problema dell’atterraggio autonomo aria-aria (Autonomous Airto-Air Landing, AAAL), in cui un piccolo drone (UAV) atterra autonomamente su un drone portaerei di dimensioni maggiori in movimento. Questa manovra complessa richiede l’integrazione della navigazione basata sulla visione, la stima dello stato e il controllo in condizioni dinamiche. Basandosi su precedenti ricerche condotte presso il Politecnico di Milano, in questa tesi vengono proposti tre contributi principali. In primo luogo, è stato implementato un filtro di Stima Adattiva dell’Ingresso e dello Stato (Adaptive Input and State Estimation, AISE) come alternativa al convenzionale Filtro di Kalman (KF). A differenza del KF, il filtro AISE stima simultaneamente sia gli stati di velocità che quelli di accelerazione. In secondo luogo, sono stati sviluppati due compensatori armonici: uno Standard (Standard Harmonic Compensator, SHC) e uno Adattivo (Adaptive Harmonic Compensator, AHC), con l’obiettivo di migliorare la robustezza contro disturbi oscillatori e garantire un preciso inseguimento di traiettorie armoniche. In terzo luogo, è stato progettato un ambiente di simulazione in Simulink utilizzando l’interfaccia Unreal Engine, che consente una visualizzazione realistica di scenari di atterraggio 3D, inclusa la rilevazione di marker ArUco basata su telecamera e le interazioni dinamiche tra UAV. Le architetture di stima e controllo proposte sono state ampiamente valutate tramite simulazioni che coinvolgono diverse traiettorie del drone carrier, dalle rettilinee a quelle più complesse di tipo circolare e a otto. I risultati indicano che il Filtro di Kalman supera l’AISE nella stima della velocità, permettendo atterraggi più rapidi e stabili. Tra i compensatori, lo SHC ha mostrato la massima robustezza, riuscendo ad atterrare anche a velocità angolari elevate del portaerei. Il framework di simulazione sviluppato si è dimostrato una piattaforma efficace e versatile per testare strategie di stima e controllo basate sulla visione nelle operazioni di AAAL. Infine, esperimenti in volo hanno validato le prestazioni superiori dei compensatori proposti rispetto ai precedenti controllori adattivi, confermando la robustezza dello SHC e l’adattabilità dell’AHC. In sintesi, questo lavoro contribuisce all’avanzamento delle tecnologie di atterraggio autonomo per UAV introducendo metodi migliorati di stima e rifiuto dei disturbi e fornendo un ambiente di simulazione validato che colma il divario tra teoria e implementazione sperimentale.
Adaptive vision-based control for autonomous air-to-air landing of multi-rotor UAVs
Perrinó Gamino, Álvaro;Revoltos Salmador, Alex
2024/2025
Abstract
This thesis investigates the problem of Autonomous Air-to-Air Landing (AAAL), in which a small unmanned aerial vehicle (UAV) autonomously lands on a larger carrier drone in motion. This complex task integrates vision-based navigation, state estimation, and control under dynamic conditions. Building upon previous research conducted at the Politecnico di Milano, three main contributions are presented. First, an Adaptive Input and State Estimation (AISE) filter is implemented as an alternative to the conventional Kalman Filter (KF). Unlike the KF, the AISE filter simultaneously estimates both velocity and acceleration states. Second, two harmonic disturbance compensators are developed, a Standard Harmonic Compensator (SHC) and an Adaptive Harmonic Compensator (AHC), to improve robustness against oscillatory disturbances and ensure accurate tracking of harmonic trajectories. Third, a simulation environment is designed in Simulink using the Unreal Engine interface, enabling realistic visualization of 3D landing scenarios, including camera-based ArUco marker detection and dynamic interactions between UAVs. The proposed estimation and control architectures are extensively evaluated through simulations involving various carrier trajectories, from rectilinear to complex circular and figure-eight paths. Results indicate that the Kalman Filter outperforms the AISE in velocity estimation, resulting in faster and more stable landings. Among the compensators, the SHC exhibited the highest robustness, achieving successful landings even at increased carrier angular velocities. The developed simulation framework proved to be an effective and versatile platform for testing vision-based estimation and control strategies in AAAL operations. Finally, in-flight experiments validated the superior performance of the proposed compensators compared to previous adaptive controllers, confirming the robustness of the SHC and the adaptability of the AHC. Overall, this work advances the development of autonomous landing technologies for UAVs by introducing enhanced estimation and disturbance rejection methods and by providing a validated simulation environment that bridges the gap between theory and experimental implementation.| File | Dimensione | Formato | |
|---|---|---|---|
|
2025_12_Revoltos_Perrino_Executive_Summary.pdf
accessibile in internet per tutti
Descrizione: Executive Summary
Dimensione
1.15 MB
Formato
Adobe PDF
|
1.15 MB | Adobe PDF | Visualizza/Apri |
|
2025_12_Revoltos_Perrino_Thesis.pdf
accessibile in internet per tutti
Descrizione: Thesis
Dimensione
18.83 MB
Formato
Adobe PDF
|
18.83 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/10589/247225