In the last few years, we are experiencing an exponential proliferation of research and applications concerning robotics and machine vision. Robots are required to autonomously operate based on information derived from sensors or vision systems. Due to the variability in this data, it is mandatory to test the safety, efficiency, and robustness of any new algorithm by means of realistic and reliable simulations. In parallel, there is still a lack of a product that can accelerate and improve the training of students and non-specialized industrial workers. Trying to address all these issues in a unique solution, we introduce a free powerful simulation software, called Vostok, that allows interacting with robots, vision systems and sensors within a dynamic operating environment. Vostok aims to help manufacturers, developers, researchers and students by providing a guided and user friendly interface through which to design, program and test autonomous machines. This work aims to achieve three objectives. 1. Provide a totally free solution that encourages individuals who want to experiment with robotics but fail due to high hardware and software costs. 2. Allow companies testing their projects within a safe environment and, consequently, to cut down long installation time. This enables developers to deploy the automation process faster and more confidently. 3. Exploit deep learning techniques at industrial level. One of the main problems regards the availability of data to train those systems. With this aim in mind, we would like to be able of generating datasets of synthetic images. The work presented in this manuscript was carried out in collaboration with Euclid Labs, who shared their ideas, experience and provided the necessary tools to validate the results.
Negli ultimi anni, stiamo assistendo a una diffusione esponenziale di ricerche e applicazioni riguardanti la robotica e la visione artificiale. I robot devono operare autonomamente sulla base di informazioni derivanti da sensori o sistemi di visione. A causa della variabilità di questi dati, è obbligatorio testare la sicurezza, l’efficienza e la robustezza di qualsiasi nuovo algoritmo per mezzo di simulazioni realistiche e affidabili. Parallelamente, manca ancora un prodotto che permetta di accelerare e migliorare la formazione di studenti ed operatori industriali non specializzati. Cercando di affrontare tutti questi problemi in un’unica soluzione, introduciamo un potente software di simulazione, chiamato Vostok, che permette di interagire con robot, sistemi di visione e sensori in un ambiente operativo dinamico. Vostok mira ad aiutare produttori, sviluppatori, ricercatori e studenti fornendo un’interfaccia guidata e facile da usare attraverso la quale progettare, programmare e testare macchine autonome. Questo lavoro mira a raggiungere tre obiettivi. 1. Fornire una soluzione totalmente gratuita che incoraggi gli individui che vogliono sperimentare la robotica ma non riescono a causa degli alti costi di hardware e software. 2. Permettere alle aziende di testare i nuovi progetti in un ambiente sicuro e, di conseguenza, ridurre i lunghi tempi di installazione. Questo permette agli sviluppatori di implementare il processo di automazione più velocemente e in sicurezza. 3. Sfruttare le tecniche di deep learning a livello industriale. Uno dei problemi principali riguarda la disponibilità di dati per addestrare questi sistemi. Con questo obiettivo in mente, vorremmo essere in grado di generare dataset di immagini sintetiche. Il lavoro presentato in questa tesi è stato realizzato in collaborazione con Euclid Labs, che ha condiviso le sue idee, l’esperienza e ha fornito gli strumenti necessari per validare i risultati.
Accelerare lo sviluppo di macchine autonome con Vostok
ROSSI, ALESSANDRO
2022
Abstract
In the last few years, we are experiencing an exponential proliferation of research and applications concerning robotics and machine vision. Robots are required to autonomously operate based on information derived from sensors or vision systems. Due to the variability in this data, it is mandatory to test the safety, efficiency, and robustness of any new algorithm by means of realistic and reliable simulations. In parallel, there is still a lack of a product that can accelerate and improve the training of students and non-specialized industrial workers. Trying to address all these issues in a unique solution, we introduce a free powerful simulation software, called Vostok, that allows interacting with robots, vision systems and sensors within a dynamic operating environment. Vostok aims to help manufacturers, developers, researchers and students by providing a guided and user friendly interface through which to design, program and test autonomous machines. This work aims to achieve three objectives. 1. Provide a totally free solution that encourages individuals who want to experiment with robotics but fail due to high hardware and software costs. 2. Allow companies testing their projects within a safe environment and, consequently, to cut down long installation time. This enables developers to deploy the automation process faster and more confidently. 3. Exploit deep learning techniques at industrial level. One of the main problems regards the availability of data to train those systems. With this aim in mind, we would like to be able of generating datasets of synthetic images. The work presented in this manuscript was carried out in collaboration with Euclid Labs, who shared their ideas, experience and provided the necessary tools to validate the results.File | Dimensione | Formato | |
---|---|---|---|
tesi_definitiva_Alessandro_Rossi.pdf
accesso aperto
Dimensione
5.34 MB
Formato
Adobe PDF
|
5.34 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/83855
URN:NBN:IT:UNIPD-83855