This thesis work origins from the collaboration between the research group DeustoTech and the SMARTiAGO project. The goal is to deploy two autonomous vehicles that can be used for last mile delivery of construction materials and subsurface inspection using a Ground Penetrating Radar. The vehicle is a PIXLOOP Hooke, an open-source autonomous driving chassis pre-arranged with all necessary sensors needed. It is compatible with Autoware.AI, an open-source software platform for self-driving vehicles. It was requested that the vehicle be capable of drive autonomously from its position to a target one, following a previously recorded path and generating the trajectory by itself. It have to perform it while detecting, tracking and avoiding cars, persons, and obstacles in its environment. Additionally, the vehicle has to generate a full path to drive over a given area with the aim of mapping the underground through the use of a Geo radar. Initially, the vehicle underwent a thorough an examination to have a better understanding of his capabilities and limitations. This involved conducting a detailed analysis of all the hardware configuration, actuators and sensors, present on the vehicle. Once completed, the focus move to the software component. The objective was to explore the various libraries available and determine how they would be used for implementing the system functionality. After this analysis was completed, the goal was to identify the most effective ways to utilize the available resources and create a robust, reliable, and efficient system. This involved integrating the software library and correctly set up the different software component needed to perform the driving task. Once the implementation was complete, two validation tests were performed to ensure that the vehicle is capable of executing the driving task. The two validation tests were critical in ensuring that the vehicle was reliable, accurate, and efficient, and could perform the autonomous driving task with high precision and consistency. The validation tests were successfully passed in December 2022. In the second phase of this thesis, the focus shift to implementing a Complete Coverage Path Planning algorithm which was done by developing a Python software compatible with Autoware.ai. The software, given the corners of a area, generates a trajectory above it that ensures complete coverage. Although it was not possible to conduct tests in a real world scenario due to unavailability of the car for further testing after the validation phase, some local tests were carried out, and the results were promising. Overall, while local testing is an essential step, further testing in real-world scenarios is necessary to fully evaluate the car’s capabilities and limitations.
Questo lavoro di tesi nasce dalla collaborazione tra il gruppo di ricerca DeustoTech e il progetto SMARTiAGO. L'obiettivo del progetto SMARTiAGO è quello di sviluppare due veicoli autonomi che possano essere utilizzati per il "last mile delivery" di materiali da costruzione e per l'ispezione del sottosuolo utilizzando un Ground Penetrating Radar. Il veicolo utilizzato è un PIXLOOP Hooke: un telaio open-source a guida autonoma già predisposto di tutti i sensori necessari, il veicolo è compatibile con Autoware.AI, un software open-source per veicoli a guida autonoma. La richiesta da cui nasce la collaborazione è che il veicolo sia in grado di guidare autonomamente dalla sua posizione ad una posizione di target, sia seguendo un traiettoria precedentemente registrata che generandola autonomamente. Durante la guida deve essere in grado di rilevare, seguire ed evitare auto, persone e ostacoli nei dintorni. Inoltre, con l'obbiettivo di mappare il sottosuolo, il veicolo deve generare una traiettoria per coprire completamente una data area. Inizialmente il veicolo è stato sottoposto a un esame approfondito per comprendere meglio le sue capacità e i suoi limiti. Questo ha comportato un'analisi dettagliata di tutta la configurazione hardware, degli attuatori e dei sensori presenti sul veicolo; in seguito il focus è stato posto sulla componente software con l'obiettivo di approfondire le varie librerie disponibili per l'implementare le funzionalità del sistema. Finita questa fase di analisi l'obiettivo è stato quello di identificare i modi più efficaci per creare un sistema robusto, affidabile ed efficiente. Ciò ha comportato l'integrazione della libreria software e la corretta parametrizzazione dei componenti necessari per svolgere il compito di guida autonoma. Una volta completato sono stati eseguiti due test di validazione per garantire che il veicolo fosse in grado di guidare autonomamente. I due test di validazione sono stati fondamentali per garantire che il veicolo fosse affidabile, accurato ed efficiente e che potesse eseguire il compito di guida autonoma con elevata precisione e coerenza. I test sono stati superati con successo nel dicembre 2022. Nella seconda fase di questa tesi è stato implementato di un algoritmo di complete coverage path planning realizzato sviluppando un software Python compatibile con Autoware.ai. Il software, dati gli angoli di un'area, genera una traiettoria all'interno di essa che garantisca una copertura completa. Sebbene non sia stato possibile condurre test in uno scenario reale a causa dell'indisponibilità dell'auto per ulteriori prove dopo la fase di convalida, sono stati condotti alcune simulazioni in un ambiente locale e i risultati sono stati promettenti. Nel complesso, sebbene i test locali siano un passo essenziale, sono necessari ulteriori test in scenari reali per valutare appieno le capacità e i limiti dell'auto.
autonomous driving configuration of a pxiloop chassis
Troiani, Alberto
2022/2023
Abstract
This thesis work origins from the collaboration between the research group DeustoTech and the SMARTiAGO project. The goal is to deploy two autonomous vehicles that can be used for last mile delivery of construction materials and subsurface inspection using a Ground Penetrating Radar. The vehicle is a PIXLOOP Hooke, an open-source autonomous driving chassis pre-arranged with all necessary sensors needed. It is compatible with Autoware.AI, an open-source software platform for self-driving vehicles. It was requested that the vehicle be capable of drive autonomously from its position to a target one, following a previously recorded path and generating the trajectory by itself. It have to perform it while detecting, tracking and avoiding cars, persons, and obstacles in its environment. Additionally, the vehicle has to generate a full path to drive over a given area with the aim of mapping the underground through the use of a Geo radar. Initially, the vehicle underwent a thorough an examination to have a better understanding of his capabilities and limitations. This involved conducting a detailed analysis of all the hardware configuration, actuators and sensors, present on the vehicle. Once completed, the focus move to the software component. The objective was to explore the various libraries available and determine how they would be used for implementing the system functionality. After this analysis was completed, the goal was to identify the most effective ways to utilize the available resources and create a robust, reliable, and efficient system. This involved integrating the software library and correctly set up the different software component needed to perform the driving task. Once the implementation was complete, two validation tests were performed to ensure that the vehicle is capable of executing the driving task. The two validation tests were critical in ensuring that the vehicle was reliable, accurate, and efficient, and could perform the autonomous driving task with high precision and consistency. The validation tests were successfully passed in December 2022. In the second phase of this thesis, the focus shift to implementing a Complete Coverage Path Planning algorithm which was done by developing a Python software compatible with Autoware.ai. The software, given the corners of a area, generates a trajectory above it that ensures complete coverage. Although it was not possible to conduct tests in a real world scenario due to unavailability of the car for further testing after the validation phase, some local tests were carried out, and the results were promising. Overall, while local testing is an essential step, further testing in real-world scenarios is necessary to fully evaluate the car’s capabilities and limitations.| File | Dimensione | Formato | |
|---|---|---|---|
|
Thesis_Troiani_Alberto.pdf
accessibile in internet solo dagli utenti autorizzati
Dimensione
14.01 MB
Formato
Adobe PDF
|
14.01 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/10589/204525