License: | Title: | Stabilisierung unkontrollierter Flugzustände mit Reinforcement Learning | Other Titles: | Stabilization of uncontrolled flight states with reinforcement learning | Language: | German | Authors: | Rohden, Andre | Keywords: | reinforcement learning; deep deterministic policy gradient; experience replay memory; curriculum learning; quadcopter | Issue Date: | 17-Apr-2019 | Abstract: | Reinforcement Learning ermöglicht einem selbstlernenden Agenten ein unbemanntes Flugobjekt in unkontrollierten Flugzuständen zu stabilisieren. Um dies zu erreichen, wird ein Deep Deterministic Policy Gradient Algorithmus angewendet. Durch Erweiterung wie Experience Replay Speicher, parametrisiertem Rauschen, Prioritized Experience Replay, Hindsight Experience Replay und Curriculum Learning lassen sich darüberhinaus Umgegebung mit sparse Reward trainieren. Reinforcement learning allows a self-learning agent to stabilize an unmanned aerial vehicle in uncontrolled flight states. To achieve this, a deep deterministic policy gradient algorithm is applied. Through extensions like experience replay memory, parameterized noise, prioritized experience replay, hindsight experience replay and curriculum learning, it is furthermore possible to train environments with sparse reward. |
URI: | http://hdl.handle.net/20.500.12738/8674 | Institute: | Department Informatik | Type: | Thesis | Thesis type: | Master Thesis | Advisor: | Meisel, Andreas | Referee: | Fohl, Wolfgang |
Appears in Collections: | Theses |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
masterthesis_rohden.pdf | 18.29 MB | Adobe PDF | View/Open |
Note about this record
Export
Items in REPOSIT are protected by copyright, with all rights reserved, unless otherwise indicated.