Strojno učenje i robotika

undergraduate thesis

• Title (english): Machine Learning and Robotics
• Author: Ivan Bilić
• Mentor: Vladimir Čeperić (mentor)
• Committee member: Vladimir Čeperić (predsjednik povjerenstva)
• Committee member: Ivan Filković (član povjerenstva)
• Committee member: Alma Traživuk (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2019-07-12, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Electrical Engineering

Abstract

Robotski operativni sustav je skup programskih okvira koji omogućuju razvoj programske podrške za robotske sustave. Sastoji se od skupa nezavisnih komponenata razvijenih od strane inženjera i istraživača. Implementiran je kao distribuirani sustav s mehanizmom objave/pretplate, u kojem je svaki program čvor, što se prikazuje ROS grafom. Nadalje, moguće ga je integrirati s naprednim tehnologijama s područja strojnog učenja što otvara vrata novim istraživanjima i spoznajama. Potporno učenje je grana strojnog učenja koja se formalizira koristeći model Markovljevog procesa odlučivanja. Fundamentalne karakteristike potpornog učenja su interaktivno učenje kroz pokušaj i pogrešku (učenje kroz iskustvo) te cilj u obliku maksimizacije kumulativne sume nagrada uz odabrani faktor propadanja. Velik izazov u potpornom učenju je i problem ravnoteže između istraživanja i eksploatacije.

Abstract (english)

Robotic operating system is a collection of software frameworks for robot software development. ROS is composed from a set of independent components that are developed and maintained by various engineers and researchers. It is implemented as a distributed system along with the publish/subscribe mechanism, where every program is a node, which is visualized by the ROS graph. Furthermore, integration of ROS with advanced machine learning technologies is also possible, which brings new opportunities for research and cognition. Reinforcement learning is a machine learning sub-area that is formalized using the Markov decision process model. Fundamental reinforcement learning characteristics include interactive learning by trial and error (learning by experience) and a goal in the shape of maximizing the discounted cumulative sum of rewards. Also, one of the great challenges of reinforcement learning is the balance between exploration and exploitation.

• Keywords: Robotski operativni sustav; distribuirani sustavi; potporno učenje; umjetna inteligencija; učenje s pokušajem i pogreškom
• Keywords (english): Robotic operating system; distributed systems; reinforcement learning; artificial intelligence; trial-and-error learning
• Language: croatian
• URN:NBN: urn:nbn:hr:168:777710
• Study programme: Title: Electrical Engineering and Information Technology; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka elektrotehnike i informacijske tehnologije (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka elektrotehnike i informacijske tehnologije)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2020-01-07 16:47:22