Autonomous vehicles are a rising technology that has recently attracted much interest. They have found many applications: from classic cars to industrial tools, but also shuttles for closed environments, like hospitals, airports, and campuses. In this work we focus on the last kind of vehicles. These operate much differently from regular cars since, in this environment, pedestrians can roam freely, and their movement is not easily predictable. For this reason, shuttles have to move at a much lower speed and consider obstacles much more carefully. Furthermore, in these environments, the GPS is absent or not always present, so it can not be relied upon. For the same reasons, the proposed solution must rely on different means. In this work, we develop multiple pipelines for mapping, localization, and navigation of autonomous shuttles in unstructured environments. We then test these on an urban shuttle inside a university campus. Using the available data, we compare the various solutions to establish the best one for our setup. In particular, we compare three state-of-the-art solutions based on a 2D model, with one that uses a 3D approach, to analyze how this impacts the performance.
I veicoli autonomi sono una tecnologia emergente che ha recentemente attirato molto interesse. Hanno trovato molte applicazioni: dalle auto classiche agli strumenti industriali, ma anche navette per ambienti chiusi, come ospedali, aeroporti e campus. In questo lavoro ci concentriamo sull'ultimo tipo di veicoli. Questi funzionano in modo molto diverso dalle normali auto poiché, in questo ambiente, i pedoni possono spostarsi liberamente e il loro movimento non è facilmente prevedibile. Per questo motivo, le navette devono muoversi a una velocità molto inferiore e considerare gli ostacoli con molta più attenzione. Inoltre, in questi ambienti, il GPS è assente o non sempre presente, quindi non ci si può fidare. Per le stesse ragioni, la soluzione proposta deve basarsi su mezzi diversi. In questo lavoro, sviluppiamo più pipeline per la mappatura, la localizzazione e la navigazione di navette autonome in ambienti non strutturati. Quindi li testiamo su una navetta urbana all'interno di un campus universitario. Utilizzando i dati disponibili, confrontiamo le varie soluzioni per stabilire quella migliore per il nostro setup. In particolare, confrontiamo tre soluzioni dello stato dell'arte basate su un modello 2D, con una che utilizza un approccio 3D, per analizzare come ciò influisca sulle prestazioni.
Mapping, localization, and navigation of an autonomous urban shuttle
Gadola, Alessandro
2021/2022
Abstract
Autonomous vehicles are a rising technology that has recently attracted much interest. They have found many applications: from classic cars to industrial tools, but also shuttles for closed environments, like hospitals, airports, and campuses. In this work we focus on the last kind of vehicles. These operate much differently from regular cars since, in this environment, pedestrians can roam freely, and their movement is not easily predictable. For this reason, shuttles have to move at a much lower speed and consider obstacles much more carefully. Furthermore, in these environments, the GPS is absent or not always present, so it can not be relied upon. For the same reasons, the proposed solution must rely on different means. In this work, we develop multiple pipelines for mapping, localization, and navigation of autonomous shuttles in unstructured environments. We then test these on an urban shuttle inside a university campus. Using the available data, we compare the various solutions to establish the best one for our setup. In particular, we compare three state-of-the-art solutions based on a 2D model, with one that uses a 3D approach, to analyze how this impacts the performance.File | Dimensione | Formato | |
---|---|---|---|
Thesis.pdf
accessibile in internet per tutti
Descrizione: Tesi
Dimensione
18 MB
Formato
Adobe PDF
|
18 MB | Adobe PDF | Visualizza/Apri |
exec.pdf
accessibile in internet per tutti
Descrizione: Executive Summary
Dimensione
5.61 MB
Formato
Adobe PDF
|
5.61 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/10589/196675