Please use this identifier to cite or link to this item: http://dx.doi.org/10.14279/depositonce-8307
Main Title: State Representation Learning with Robotic Priors for Partially Observable Environments Data
Author(s): Morik, Marco
Rastogi, Divyam
Brock, Oliver
Type: Generic Research Data
Language Code: en
Abstract: We introduce Recurrent State Representation Learning (RSRL) to tackle the problem of state representation learning in robotics for partially observable environments. To learn low dimensional state representations, we combine a Long Short Term Memory network with robotic priors. RSRL introduces new priors with landmarks and combines them with existing robotics priors in literature to train the representations. To evaluate the quality of the learned state representation, we introduce validation networks that help us to better visualize and quantitatively analyze the learned state representations. We show that the learned representations are low dimensional, locally consistent, and can approximate the underlying true state for robot localization in simulated 3D maze environments. We use the learned representations for reinforcement learning and show that we achieve similar performance as training with the true state. The learned representations are also robust to landmark misclassification errors.
URI: https://depositonce.tu-berlin.de/handle/11303/9224
http://dx.doi.org/10.14279/depositonce-8307
Issue Date: 2019
Date Available: 23-Apr-2019
DDC Class: 500 Naturwissenschaften und Mathematik
Subject(s): Maza Data
robotic
Sponsor/Funder: DFG/329426068/Maschinelles Lernen für Probleme in der Robotik/R-ML
License: https://choosealicense.com/licenses/gpl-3.0/
Appears in Collections:FG Robotics » Research Data

Files in This Item:
File Description SizeFormat 
nav03short_train.npz809.38 MBUnknownView/Open
nav03short_test.npz809.38 MBUnknownView/Open
nav02short_train.npz809.38 MBUnknownView/Open
nav02short_test.npz809.38 MBUnknownView/Open
nav03_train.npz404.69 MBUnknownView/Open
nav03_test.npz404.69 MBUnknownView/Open
nav02_train.npz404.69 MBUnknownView/Open
nav02_test.npz404.69 MBUnknownView/Open
nav01short_train.npz809.38 MBUnknownView/Open
nav01short_test.npz809.38 MBUnknownView/Open
nav01_test.npz404.69 MBUnknownView/Open
nav01_train.npz404.69 MBUnknownView/Open
data_utils.py12.55 kBUnknownView/Open
README.md684 BUnknownView/Open
Masterthesis_RSRL.pdf7.64 MBAdobe PDFThumbnail
View/Open
data_utils.py12.55 kBUnknownView/Open


Items in DepositOnce are protected by copyright, with all rights reserved, unless otherwise indicated.