In this project, it is discussed the autonomous navigation of a two-wheeled vehicle in pedestrian areas, in the context of parcels delivery in an urban environment. In the preliminary phase, the problems that a vehicle may encounter during navigation in an urban environment will be presented. In this regard, a quantitative index will be proposed to evaluate in semi-automatic way the feasibility of a path on sidewalks. An analysis of the sensors that can be used for the localization and perception of the environment around the robot will also be presented. Particular attention will be given to the main GNSS technologies, showing that, due to the high uncertainty and dependence on sky coverage conditions, they can not be used for localization without the aid of other sensors. Without the information on the environment to be visited and an instrument for a precise localization, the problem of autonomous driving is defined as SLAM (Simultaneus Localization And Mapping), i.e. the robot must at the same time create a map and locate itself within it. The main paradigms to solve this problem and the considerations made in choosing the best for the use case of this project will then be described. The SLAM algorithm was implemented, offline, for the creation of the map. The choice of this approach was guided by the need to obtain a consistent and accurate map, which required great computational effort and therefore the renunciation of real-time. The software for localization and navigation has been implemented inside the ROS framework.The use of the Navigation Stak package for navigation in an urban environment will be discussed, showing its potential and limits. To conclude, the results and experiences gained will be discussed.
In questo lavoro si discuter´a della navigazione autonoma di un veicolo a due ruote in zone pedonali, nell’ambito della consegna di pacchi in un ambiente urbano. Nella fase preliminare si presenteranno le problematiche che un veicolo pu´o incontrare durante la navigazione in ambiente urbano. Si proporr´a a tale proposito un indice quantitativo per valutare in modo semi-automatico la fattibilit´a di un percorso su marciapiedi. Verr´a inoltre presentata una analisi dei sensori che possono essere impiegati per la localizzazione e percezione dell’ambiente attorno al robot. Particolare attenzione sar´a data alle principali tecnologie GNSS, mostrando che, a causa della elevata incertezza e dipendenza dalle condizioni di copertura aerea, non possono essere utilizzate per la localizzazione senza l’ausilio di altri sensori. Senza l’informazione sull’ambiente da visitare e uno strumento per una precisa localizzazione il problema della guida autonoma viene definito come SLAM (Simultaneus Localization And Mapping), cio`e il robot deve allo stesso tempo creare una mappa e localizzarsi all’interno di essa. Saranno quindi descritti i principali paradigmi per risolvere questo problema e le considerazioni fatte per la scelta dell’algoritmo implementato. L’algoritmo SLAM `e stato implementato,offline, per la sola creazione della mappa. La scelta di questo approccio `e stata guidata dalla volont´a di ottenere una mappa consistente e accurata, che ha richiesto maggiori tempi di processing e quindi la rinuncia del real-time. Il software per la localizzazione e navigazione `e stato implementato con l’ausilio della piattaforma ROS, per tanto verr´a discusso l’uso del pacchetto Navigation Stak per la navigazione in ambiente urbano, mostrandone potenzialit´a e limiti. Per concludere verranno mostrati i risultati ottenuti e si discuteranno le esperienze acquisite.
Autonomous navigation in pedestrian urban areas
PIZZOCARO, SOLOMON;MUSCOLINO, LORENZO
2017/2018
Abstract
In this project, it is discussed the autonomous navigation of a two-wheeled vehicle in pedestrian areas, in the context of parcels delivery in an urban environment. In the preliminary phase, the problems that a vehicle may encounter during navigation in an urban environment will be presented. In this regard, a quantitative index will be proposed to evaluate in semi-automatic way the feasibility of a path on sidewalks. An analysis of the sensors that can be used for the localization and perception of the environment around the robot will also be presented. Particular attention will be given to the main GNSS technologies, showing that, due to the high uncertainty and dependence on sky coverage conditions, they can not be used for localization without the aid of other sensors. Without the information on the environment to be visited and an instrument for a precise localization, the problem of autonomous driving is defined as SLAM (Simultaneus Localization And Mapping), i.e. the robot must at the same time create a map and locate itself within it. The main paradigms to solve this problem and the considerations made in choosing the best for the use case of this project will then be described. The SLAM algorithm was implemented, offline, for the creation of the map. The choice of this approach was guided by the need to obtain a consistent and accurate map, which required great computational effort and therefore the renunciation of real-time. The software for localization and navigation has been implemented inside the ROS framework.The use of the Navigation Stak package for navigation in an urban environment will be discussed, showing its potential and limits. To conclude, the results and experiences gained will be discussed.File | Dimensione | Formato | |
---|---|---|---|
2018_12_Pizzocaro_Muscolino.pdf
non accessibile
Descrizione: Thesis text
Dimensione
22.54 MB
Formato
Adobe PDF
|
22.54 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/10589/144875