Self-driving vehicles are one of the revolutions set to change mobility in the coming years. One of the biggest challenges that need to be solved in this context is to obtain highly accurate localization. Current solutions relying on Global Navigation Satellite Systems (GNSSs) suffer from shadowing and blockage conditions in urban environments and cannot meet the stringent requirements foreseen for future mobility systems. Vehicle-to- Everything (V2X) communication technologies lay the foundation for the exchange of massive amounts of data among any interconnected road entity and have been demonstrated to bring several advantages, especially for increasing the localization accuracy in vehicular networks. This thesis proposes a localization algorithm aimed at increasing the positioning performance for autonomous driving applications through Vehicle-to- Infrastructure (V2I) communication. The idea is to take advantage of the vehicles’ on-board lidar sensor and a neural network, namely a 3D object detection network, to detect poles in the environment. In turn, the poles are equipped with a device, called a Smart Tag, which is able to establish a two-way communication channel with the vehicles to provide the exact location of the pole on which it is placed. At that point, the vehicles, can compute the distance to the pole and, knowing the true position and the estimated position of the pole in question, are able to obtain a difference between the measurements that they use to refine their location. The proposed localization method is evaluated in a challenging simulated automotive scenario, with complex driving behaviours, interactions among vehicles and traffic regulating systems. The analysis, carried out assuming that the poles in the scenario are equipped with with Smart Tags, shows the benefits of the developed system in achieving highly-accurate vehicle location outcomes and studies the performances by varying the Smart Tags density taking into account also environmental constraints. By allowing a more accurate estimation of the vehicles’ location, the proposed system facilities the development of several automotive use cases that have been thoroughly analyzed in the thesis.
I veicoli a guida autonoma sono una delle rivoluzioni destinate a cambiare la mobilità nei prossimi anni. Una delle maggiori sfide che devono essere risolte in questo contesto è quella di ottenere una localizzazione altamente accurata. Le attuali soluzioni che si basano sui sistemi globali di navigazione satellitare (GNSS) soffrono di condizioni di shadowing e il segnale può subire interferenze in ambienti urbani, motivo per il quale non possono soddisfare gli stringenti requisiti previsti per i futuri sistemi di mobilità. Le tecnologie di comunicazione Vehicle-to-Everything(V2X) gettano le basi per lo scambio di enormi quantità di dati tra qualsiasi entità stradale interconnessa e hanno dimostrato di portare diversi vantaggi, soprattutto al fine di aumentare l’accuratezza della localizzazione nelle reti veicolari. Questa tesi propone un algoritmo di localizzazione volto ad aumentare le prestazioni di posizionamento per le applicazioni di guida autonoma attraverso la comunicazione Vehicle-to-Infrastructure (V2I). L’idea è di sfruttare il sensore lidar a bordo dei veicoli e una rete neurale, ovvero una rete di rilevamento di oggetti 3D, per individuare i pali presenti nell’ambiente. A loro volta, i pali vengono equipaggiati con un dispositivo, denominato Smart Tag, il quale è in grado di stabilire un canale di comunicazione bidirezionale con i veicoli al fine di fornire la posizione esatta del palo su cui sono posti. A quel punto, i veicoli, possono calcolare la distanza dal palo e, conoscendo la posizione vera e la posizione stimata del palo in questione, sono in grado di ottenere una differenza tra le misure che utilizzano per perfezionare la loro localizzazione. Il metodo di localizzazione proposto viene valutato in uno scenario automobilistico simulato e impegnativo, con comportamenti di guida complessi, interazioni tra veicoli e sistemi di regolazione del traffico. L’analisi, effettuata ipotizzando che i pali presenti nello scenario siano muniti di Smart Tag, mostra i vantaggi del sistema sviluppato per ottenere risultati altamente accurati nella localizzazione dei veicoli e studia le prestazioni al variare della densità degli Smart Tag, tenendo conto anche dei vincoli ambientali. Consentendo una stima più accurata della posizione dei veicoli, il sistema proposto facilita lo sviluppo di diversi casi d’uso nel settore automobilistico che sono stati analizzati a fondo nella tesi.
Smart Tag : active marker for localization of connected automated vehicles
POSSENTI, DAVIDE
2020/2021
Abstract
Self-driving vehicles are one of the revolutions set to change mobility in the coming years. One of the biggest challenges that need to be solved in this context is to obtain highly accurate localization. Current solutions relying on Global Navigation Satellite Systems (GNSSs) suffer from shadowing and blockage conditions in urban environments and cannot meet the stringent requirements foreseen for future mobility systems. Vehicle-to- Everything (V2X) communication technologies lay the foundation for the exchange of massive amounts of data among any interconnected road entity and have been demonstrated to bring several advantages, especially for increasing the localization accuracy in vehicular networks. This thesis proposes a localization algorithm aimed at increasing the positioning performance for autonomous driving applications through Vehicle-to- Infrastructure (V2I) communication. The idea is to take advantage of the vehicles’ on-board lidar sensor and a neural network, namely a 3D object detection network, to detect poles in the environment. In turn, the poles are equipped with a device, called a Smart Tag, which is able to establish a two-way communication channel with the vehicles to provide the exact location of the pole on which it is placed. At that point, the vehicles, can compute the distance to the pole and, knowing the true position and the estimated position of the pole in question, are able to obtain a difference between the measurements that they use to refine their location. The proposed localization method is evaluated in a challenging simulated automotive scenario, with complex driving behaviours, interactions among vehicles and traffic regulating systems. The analysis, carried out assuming that the poles in the scenario are equipped with with Smart Tags, shows the benefits of the developed system in achieving highly-accurate vehicle location outcomes and studies the performances by varying the Smart Tags density taking into account also environmental constraints. By allowing a more accurate estimation of the vehicles’ location, the proposed system facilities the development of several automotive use cases that have been thoroughly analyzed in the thesis.| File | Dimensione | Formato | |
|---|---|---|---|
|
Master_Thesis_Davide_Possenti.pdf
non accessibile
Dimensione
5.1 MB
Formato
Adobe PDF
|
5.1 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/10589/189035