A Deep-Reinforcement-Learning Approach to the Peg-in-Hole Task with Goal Uncertainties

The thesis proposed a framework to train deep-reinforcement-learning agents for fine manipulation tasks with goal uncertainties. It consisted of three aspects: state-space formulation, artificial training-goals uncertainties, and progressive training. The framework was used in a simulation for two f...

Full description

Bibliographic Details
Main Author: Rouillard, Thibault
Format: Thesis
Published: Curtin University 2020
Online Access:http://hdl.handle.net/20.500.11937/85887
_version_ 1848764776110358528
author Rouillard, Thibault
author_facet Rouillard, Thibault
author_sort Rouillard, Thibault
building Curtin Institutional Repository
collection Online Access
description The thesis proposed a framework to train deep-reinforcement-learning agents for fine manipulation tasks with goal uncertainties. It consisted of three aspects: state-space formulation, artificial training-goals uncertainties, and progressive training. The framework was used in a simulation for two fine manipulation tasks, square Peg-in-Hole and round Peg-in-Hole. The resulting behaviours were then transferred a physical robotic manipulator and compared to traditional training methods. The deep-reinforcement-learning agents trained using the framework in this work outperformed those trained with definite goals.
first_indexed 2025-11-14T11:24:43Z
format Thesis
id curtin-20.500.11937-85887
institution Curtin University Malaysia
institution_category Local University
last_indexed 2025-11-14T11:24:43Z
publishDate 2020
publisher Curtin University
recordtype eprints
repository_type Digital Repository
spelling curtin-20.500.11937-858872021-10-05T05:39:11Z A Deep-Reinforcement-Learning Approach to the Peg-in-Hole Task with Goal Uncertainties Rouillard, Thibault The thesis proposed a framework to train deep-reinforcement-learning agents for fine manipulation tasks with goal uncertainties. It consisted of three aspects: state-space formulation, artificial training-goals uncertainties, and progressive training. The framework was used in a simulation for two fine manipulation tasks, square Peg-in-Hole and round Peg-in-Hole. The resulting behaviours were then transferred a physical robotic manipulator and compared to traditional training methods. The deep-reinforcement-learning agents trained using the framework in this work outperformed those trained with definite goals. 2020 Thesis http://hdl.handle.net/20.500.11937/85887 Curtin University fulltext
spellingShingle Rouillard, Thibault
A Deep-Reinforcement-Learning Approach to the Peg-in-Hole Task with Goal Uncertainties
title A Deep-Reinforcement-Learning Approach to the Peg-in-Hole Task with Goal Uncertainties
title_full A Deep-Reinforcement-Learning Approach to the Peg-in-Hole Task with Goal Uncertainties
title_fullStr A Deep-Reinforcement-Learning Approach to the Peg-in-Hole Task with Goal Uncertainties
title_full_unstemmed A Deep-Reinforcement-Learning Approach to the Peg-in-Hole Task with Goal Uncertainties
title_short A Deep-Reinforcement-Learning Approach to the Peg-in-Hole Task with Goal Uncertainties
title_sort deep-reinforcement-learning approach to the peg-in-hole task with goal uncertainties
url http://hdl.handle.net/20.500.11937/85887