This paper presents a methodology to validate Traffic Sign Recognition with automatic tests in a virtual environment, considering the heterogeneity of road signs in different countries. General concepts of automotive electrical systems are provided, an overview of Advanced Driver Assistance Systems (ADAS), notes on validation regulations, presentation of various suppliers of virtual validation tools and the two main road sign regulations in force in the world. The methodology includes the creation of a Python plugin for the VST automatic testing tool, developed internally at Stellantis, to dynamically generate virtual scenarios with the road signs of the countries of interest for validation via dSpace ModelDesk. The activity also includes the development of a database of 3D models of road signs in Italy, Sweden, USA, Japan and the United Kingdom; nations chosen for their peculiarities in terms of driving side, measurement system and pictorial representation of road signs in order to verify the behavior of the TSR in different markets. For the experimental verification of the methodology, an electrical bench representing the entire low voltage electrical system of a car is used, a Hardware in Loop (HiL) dSpace simulates the signals necessary for driving and the front camera of the vehicle frames a screen with the virtual scenario generated while running. The verification of the correct functioning of the TSR is carried out by VST by analyzing the data emitted by the respective control unit and present on the CAN bus. The bench configured in this way can operate 24 hours a day with requests that can be started remotely, return the results via e-mail and can be consulted with the browser. In the final considerations, the usefulness of the methodology is underlined in order to limit the use of human resources in low added value activities, saving time in the search for anomalies during software development and consequently in a better quality of the final product. Considerations for possible future developments of the methodology are also included.
In questo elaborato si presenta una metodologia per validare con test automatici il Traffic Sign Recognition in un ambiente virtuale, considerando l’eterogeneità dei cartelli stradali in diverse nazioni. Si forniscono concetti generali degli impianti elettrici automotive, una panoramica sui sistemi avanzati di guida assistita (ADAS), cenni sulla normativa per la validazione, presentazione di diversi fornitori di strumenti per la validazione virtuale e le due principali normative sui cartelli stradali in vigore nel mondo. La metodologia comprende la realizzazione un plugin Python per il tool di test automatici VST, sviluppato internamente in Stellantis, per generare dinamicamente tramite dSpace ModelDesk gli scenari virtuali con i cartelli stradali delle nazioni d’interesse per la validazione. L’attività include anche lo sviluppo di un database di modelli 3D dei cartelli stradali di Italia, Svezia, USA, Giappone e Regno Unito; nazioni scelte per le loro peculiarità in termini di lato di guida, sistema di misura e rappresentazione pittorica dei cartelli stradali al fine di verificare il comportamento del TSR in mercati differenti. Per la verifica sperimentale della metodologia si impiega un banco elettrico rappresentante l’intero impianto elettrico a bassa tensione di un’autovettura, un Hardware in Loop (HiL) dSpace simula i segnali necessari per la marcia e la camera frontale del veicolo inquadra uno schermo con lo scenario virtuale generato in esecuzione. La verifica del corretto funzionamento del TSR viene effettuata da VST analizzando i dati emessi dalla rispettiva centralina e presenti sul bus CAN, il banco così configurato può operare 24 ore su 24 con richieste avviabili da remoto, restituire i risultati via e-mail e consultabili con il browser. Nelle considerazioni finali, si sottolinea l’utilità della metodologia al fine di limitare l’impiego di risorse umane in attività a basso valore aggiunto, il risparmio di tempo nella ricerca delle anomalie durante lo sviluppo del software e di conseguenza in una migliore qualità del prodotto finale. Si includono anche considerazioni per eventuali sviluppi futuri della metodologia.
Test automatici HIL per la validazione internazionale del Traffic Sign Recognition
ANGELINI, MATTEO
2022/2023
Abstract
This paper presents a methodology to validate Traffic Sign Recognition with automatic tests in a virtual environment, considering the heterogeneity of road signs in different countries. General concepts of automotive electrical systems are provided, an overview of Advanced Driver Assistance Systems (ADAS), notes on validation regulations, presentation of various suppliers of virtual validation tools and the two main road sign regulations in force in the world. The methodology includes the creation of a Python plugin for the VST automatic testing tool, developed internally at Stellantis, to dynamically generate virtual scenarios with the road signs of the countries of interest for validation via dSpace ModelDesk. The activity also includes the development of a database of 3D models of road signs in Italy, Sweden, USA, Japan and the United Kingdom; nations chosen for their peculiarities in terms of driving side, measurement system and pictorial representation of road signs in order to verify the behavior of the TSR in different markets. For the experimental verification of the methodology, an electrical bench representing the entire low voltage electrical system of a car is used, a Hardware in Loop (HiL) dSpace simulates the signals necessary for driving and the front camera of the vehicle frames a screen with the virtual scenario generated while running. The verification of the correct functioning of the TSR is carried out by VST by analyzing the data emitted by the respective control unit and present on the CAN bus. The bench configured in this way can operate 24 hours a day with requests that can be started remotely, return the results via e-mail and can be consulted with the browser. In the final considerations, the usefulness of the methodology is underlined in order to limit the use of human resources in low added value activities, saving time in the search for anomalies during software development and consequently in a better quality of the final product. Considerations for possible future developments of the methodology are also included.| File | Dimensione | Formato | |
|---|---|---|---|
|
Tesi_AngeliniMatteo_Approved.pdf
non disponibili
Dimensione
73.65 MB
Formato
Adobe PDF
|
73.65 MB | Adobe PDF |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.12075/16564