This work has been carried out within the framework of the Bosch Future Mobility Challenge (BFMC), an international competition that involves university students in the development of autonomous driving and intelligent mobility solutions. In this context, the thesis addresses the problem of vehicle localization and environment mapping, which are essential aspects for the proper operation of advanced autonomous driving functionalities. The main objective is to define and implement a methodology that, starting from sensor fusion based on nonlinear Gaussian filters, improves odometric estimation through noise reduction techniques. On this basis, the global localization phase has been developed by means of simultaneous localization and mapping (SLAM) algorithms for the construction of the environmental map. The generated map is subsequently employed by the adaptive Monte Carlo localization (AMCL) package in order to enhance the accuracy and robustness of the vehicle pose estimation. Experimental results have shown an average localization error below 0.10 m and improved system stability under dynamic conditions, confirming the effectiveness of the proposed approach. The main contribution of this work lies in the design and validation of a complete and modular localization architecture, fully integrated within the ROS 2 environment, capable of coherently combining sensor fusion, SLAM, and AMCL into a single operational framework.
Il presente elaborato si colloca nell’ambito della Bosch Future Mobility Challenge (BFMC), competizione internazionale che coinvolge studenti universitari nello svilup- po di soluzioni per la guida autonoma e la mobilità intelligente. In tale contesto, il lavoro affronta il problema della localizzazione del veicolo e della mappatura dell’am- biente di test, aspetti fondamentali per il corretto funzionamento delle funzionalità avanzate di guida autonoma. L’obiettivo è definire e implementare una metodologia che, a partire dalla fusione sensoriale basata su filtri gaussiani non lineari, migliori la stima odometrica me- diante tecniche di riduzione del rumore. Su tale base viene sviluppata la fase di localizzazione globale, realizzata attraverso algoritmi di simultaneous localization and mapping (SLAM) per la costruzione della mappa dell’ambiente. La mappa generata è successivamente impiegata dal pacchetto adaptive Monte Carlo localization (AMCL), con lo scopo di incrementare la precisione e la robustezza della stima di posa del veicolo. I risultati sperimentali hanno evidenziato una riduzione dell’errore medio di localizzazione al di sotto di 0.10 m e una maggiore stabilità del sistema in condizioni dinamiche, confermando l’efficacia dell’approccio proposto. Il contributo principale del lavoro consiste nella progettazione e validazione di un’architettura di localiz- zazione completa, modulare e pienamente integrata in ambiente ROS 2, capace di combinare in modo coerente fusione sensoriale, SLAM e AMCL in un unico framework operativo
Implementazione di sistemi per auto-localizzazione e map-building di auto a guida autonoma.
FACCENDA, ANDREA
2025/2026
Abstract
This work has been carried out within the framework of the Bosch Future Mobility Challenge (BFMC), an international competition that involves university students in the development of autonomous driving and intelligent mobility solutions. In this context, the thesis addresses the problem of vehicle localization and environment mapping, which are essential aspects for the proper operation of advanced autonomous driving functionalities. The main objective is to define and implement a methodology that, starting from sensor fusion based on nonlinear Gaussian filters, improves odometric estimation through noise reduction techniques. On this basis, the global localization phase has been developed by means of simultaneous localization and mapping (SLAM) algorithms for the construction of the environmental map. The generated map is subsequently employed by the adaptive Monte Carlo localization (AMCL) package in order to enhance the accuracy and robustness of the vehicle pose estimation. Experimental results have shown an average localization error below 0.10 m and improved system stability under dynamic conditions, confirming the effectiveness of the proposed approach. The main contribution of this work lies in the design and validation of a complete and modular localization architecture, fully integrated within the ROS 2 environment, capable of coherently combining sensor fusion, SLAM, and AMCL into a single operational framework.| File | Dimensione | Formato | |
|---|---|---|---|
|
Tesi_Andrea_Faccenda.pdf
accesso aperto
Dimensione
4.27 MB
Formato
Adobe PDF
|
4.27 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.12075/28209