Robotsko rukovanje predmetima rada u nestrukturiranoj radnoj okolini

master's thesis

• Title (english): Robotic handling of workpieces in unstructured working environment
• Author: Eugen Cicvarić
• Mentor: Bojan Jerbić (mentor)
• Granter: University of Zagreb; Faculty of Mechanical Engineering and Naval Architecture; Zagreb
• Defense date and country: 2018-12-06, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Electrical Engineering; Automation and Robotics

Abstract

U radu su razvijena i implementirana dva načina prepoznavanja te prostorne lokalizacije objekata na sceni koristeći senzor Kinect v2 i softversku pozadinu Point Cloud Library. Pri tome korišten je robot UR5 koji sa servera, putem TCP protokola, prima podatke o lokaciji i pretvara ih u naredbe gibanja. Kako bi dohvaćen oblak točaka iz senzora bio prikazan sa što više konzistentnosti potrebno je primjeniti filtraciju i algoritme za rekonstrukciju površine. Za obje metode, ekstrakcije klastera i lokalnih značajki, napravljena je teoretska pozadina i softverska implementacija na temelju koje su utvrđene prednosti, odnosno nedostaci. Za metodu lokalnih značajki napravljena su eksperimentalna ispitivanja preciznosti i prikaz rada u praktičnoj primjeni. Ovakvi pristupi prepoznavanja i lokalizacije neovisni su o geometriji promatranih objekata i moguće ih je prilagoditi za sve industrijske robote. Osim spomenutog Point Cloud Library-a korišteni su softveri CMake za izradu projekta, te Visual Studio 2015 za C++ programiranje. Struktura rada sastoji se od uvoda, osnovnih informacija o korištenim elementima sustava i njihovoj interakciji, zatim je razrađena teorija za razvijene metode, slijedi implementacija teoretske pozadine pomoću PCL-a, povezivanje s robotom i u konačnici ispitivanje preciznosti te zaključak.

Abstract (english)

Two ways of detecting and spatial localization of objects are developed and implemented in thesis using the Kinect v2 sensor and the Point Cloud Library as software background. The UR5 robot, which uses the TCP connection, receives location data and converts them into the motion instructions. In order to capture consistent points cloud from the sensor, it is necessary to apply filtration and surface reconstruction algorithms. For both methods, cluster extraction and local features, a theoretical background and software implementation were developed based on which the advantages and disadvantages were identified. For the local feature method, experimental examinations of precision and work presentation were performed in practical application. Such approaches to recognition and localization are independent of the geometry of the observed objects and can be adapted to all industrial robots. In addition to the mentioned Point Cloud Library, used softwares are also CMake and Visual Studio 2015 for C ++ programming. The structure of the work consists introduction, basic information of the system elements and their interaction, then elaborated theory for developed methods followed by the implementation using PCL, realization of connection with the robotic arm, also experimental testing about the precision and conclusion.

• Keywords: Robotska vizija; prepoznavanje; lokalizacija; Point Cloud Library; oblak točaka; UR5
• Keywords (english): Robotic vision; detection; localization; Point Cloud Library; point cloud; UR5
• Language: croatian
• URN:NBN: urn:nbn:hr:235:080050
• Study programme: Title: Mechanical Engineering; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka strojarstva (magistar/magistra inženjer/inženjerka strojarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2019-05-24 12:57:19