User: Guest  Login
If you experience problems opening the document, please try this link.
Original title:
Safe Reinforcement Learning Methods for Complex Dynamical Systems Based on Model Order Reduction Techniques
Translated title:
Sichere Reinforcement-Learning-Methoden für komplexe dynamische Systeme basierend auf Modellordnungsreduktionstechniken
Author:
Zhou, Zhehua
Year:
2022
Document type:
Dissertation
Faculty/School:
TUM School of Computation, Information and Technology
Advisor:
Buss, Martin (Prof. Dr.)
Referee:
Buss, Martin (Prof. Dr.); Diepold, Klaus (Prof. Dr.)
Language:
en
Subject group:
ELT Elektrotechnik
Keywords:
safe reinforcement learning, dynamical systems
Translated keywords:
sichere Reinforcement-Learning, dynamische Systeme
TUM classification:
MSR 600
Abstract:
Reinforcement learning is used in many recent studies for solving complicated control tasks. For safely applying reinforcement learning on real-world systems, we focus in this dissertation on proposing general safe reinforcement learning methods that are suitable for complex dynamical systems. By using a supervisory control strategy, a safe reinforcement learning framework is realized and is validated in multiple control scenarios and systems.
Translated abstract:
Reinforcement Learning wird in vielen neueren Studien zur Lösung komplizierter Steuerungsaufgaben eingesetzt. Um Reinforcement Learning sicher auf reale Systeme anzuwenden, konzentrieren wir uns in dieser Dissertation darauf, allgemeine sichere Reinforcement-Learning-Methoden vorzuschlagen. Durch die Verwendung einer Überwachungskontrollstrategie wird ein sicheres Rahmenwerk für Reinforcement Learning realisiert und in mehreren Kontrollszenarien und -systemen validiert.
WWW:
https://mediatum.ub.tum.de/?id=1652885
Date of submission:
31.03.2022
Oral examination:
24.10.2022
File size:
12968887 bytes
Pages:
141
Urn (citeable URL):
https://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:bvb:91-diss-20221024-1652885-1-6
Last change:
02.02.2023
 BibTeX