Kontroliranje dinamskega sistema s strojnim učenjem

Lackovič, David (author), Slavič, Janko (mentor)

Na področju krmiljenja dinamskih sistemov se vedno pogosteje uveljavlja strojno učenje, predvsem zaradi prilagodljivosti in zmožnosti obvladovanja kompleksnih sistemov brez poznavanja kinematike. V raziskavi te naloge je obravnavana implementacija strojnega učenja za uporabo pri usmerjanju krogle na 3-osnem paralelnem manipulatorju. Z eksperimentom je bilo ugotovljeno, da je strojno učenje primerno za tovrstno nalogo, uspešnost pa je odvisna od načina implementacije ter učenja.

• Language: Slovenian
• Keywords: dinamski sistemi, paralelni manipulatorji, strojno učenje, okrepitveno učenje, usmerjanje krogle
• Work type: Bachelor thesis/paper
• Typology: 2.11 - Undergraduate Thesis
• Organization: FS - Faculty of Mechanical Engineering
• Year: 2025
• Number of pages: XI, 28 f.
• PID: 20.500.12556/RUL-171540
• UDC: 007.52:681.52:004.85(043.2)
• COBISS.SI-ID: 247480579
• Publication date in RUL: 28.08.2025

Secondary language

• Language: English
• Title: Machine-learning based control of a dynamic system

Abstract

In the field of dynamic system control, machine learning is becoming increasingly prominent, primarily due to its flexibility and ability to handle complex systems without requiring knowledge of kinematics. Research of this thesis examines an implementation of machine learning for use in controlling ball's motion on a 3-axis parallel manipulator. With the experiment it was found that machine learning is suitable for this type of task, although its effectiveness depends on the implementation method and the training process.

• Keywords: dynamic systems, parallel manipulators, machine learning, reinforcement learning, ball balancing