State Representation Learning with Robotic Priors for Partially Observable Environments Data

dc.contributor.authorMorik, Marco
dc.contributor.authorRastogi, Divyam
dc.contributor.authorBrock, Oliver
dc.date.accessioned2019-04-23T06:59:38Z
dc.date.available2019-04-23T06:59:38Z
dc.date.issued2019
dc.description.abstractWe 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.en
dc.description.sponsorshipDFG/329426068/Maschinelles Lernen für Probleme in der Robotik/R-MLen
dc.identifier.urihttps://depositonce.tu-berlin.de/handle/11303/9224
dc.identifier.urihttp://dx.doi.org/10.14279/depositonce-8307
dc.language.isoenen
dc.relation.issupplementtohttps://doi.org/10.1109/IROS40897.2019.8967938en
dc.rights.urihttps://choosealicense.com/licenses/gpl-3.0/en
dc.subject.ddc500 Naturwissenschaften und Mathematikde
dc.subject.otherMaza Dataen
dc.subject.otherroboticen
dc.titleState Representation Learning with Robotic Priors for Partially Observable Environments Dataen
dc.typeGeneric Research Dataen
tub.accessrights.dnbunknown*
tub.affiliationFak. 4 Elektrotechnik und Informatik::Inst. Technische Informatik und Mikroelektronik::FG Roboticsde
tub.affiliation.facultyFak. 4 Elektrotechnik und Informatikde
tub.affiliation.groupFG Roboticsde
tub.affiliation.instituteInst. Technische Informatik und Mikroelektronikde

Files

Original bundle
Now showing 1 - 16 of 16
No Thumbnail Available
Name:
README.md
Size:
684 B
Format:
Unknown data format
Loading…
Thumbnail Image
Name:
Masterthesis_RSRL.pdf
Size:
7.46 MB
Format:
Adobe Portable Document Format
No Thumbnail Available
Name:
nav03short_train.npz
Size:
790.41 MB
Format:
Unknown data format
No Thumbnail Available
Name:
nav03short_test.npz
Size:
790.41 MB
Format:
Unknown data format
No Thumbnail Available
Name:
nav02short_train.npz
Size:
790.41 MB
Format:
Unknown data format
No Thumbnail Available
Name:
nav02short_test.npz
Size:
790.41 MB
Format:
Unknown data format
No Thumbnail Available
Name:
nav03_train.npz
Size:
395.2 MB
Format:
Unknown data format
No Thumbnail Available
Name:
nav03_test.npz
Size:
395.2 MB
Format:
Unknown data format
No Thumbnail Available
Name:
nav02_train.npz
Size:
395.2 MB
Format:
Unknown data format
No Thumbnail Available
Name:
nav02_test.npz
Size:
395.2 MB
Format:
Unknown data format
No Thumbnail Available
Name:
nav01short_train.npz
Size:
790.41 MB
Format:
Unknown data format
No Thumbnail Available
Name:
nav01short_test.npz
Size:
790.41 MB
Format:
Unknown data format
No Thumbnail Available
Name:
nav01_test.npz
Size:
395.2 MB
Format:
Unknown data format
No Thumbnail Available
Name:
nav01_train.npz
Size:
395.2 MB
Format:
Unknown data format
No Thumbnail Available
Name:
data_utils.py
Size:
12.55 KB
Format:
Unknown data format
No Thumbnail Available
Name:
data_utils.py
Size:
12.55 KB
Format:
Unknown data format
License bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
license.txt
Size:
2.71 KB
Format:
Item-specific license agreed upon to submission
Description:

Collections