Fraunhofer-Gesellschaft

Publica

Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Domain Shifts in Reinforcement Learning: Identifying Disturbances in Environments

Poster presented at AISafety 2021. Artificial Intelligence Safety 2021, co-located with the 30th International Joint Conference on Artificial Intelligence, IJCAI 2021, Virtual, August 2021
 
: Haider, Tom; Schmoeller Roza, Felippe; Eilers, Dirk; Roscher, Karsten; Günnemann, Stephan

:
Poster urn:nbn:de:0011-n-6404429 (224 KByte PDF)
MD5 Fingerprint: 45d6823fbcf419dcb4531d7fa22d470c
Created on: 21.9.2021


2021, 1 Folie
Workshop on Artificial Intelligence Safety (SafeAI) <2021, Online>
International Joint Conference on Artificial Intelligence (IJCAI) <30, 2021, Online>
Bayerisches Staatsministerium für Wirtschaft, Landesentwicklung und Energie StMWi

IKS-Ausbauprojekt
English
Poster, Electronic Publication
Fraunhofer IKS ()
reinforcement learning; RL; safety critical; Markov Decision Process; MDP; safety; Safe Intelligence; robustness; domain shift; out of distribution

Abstract
End-to-End Deep Reinforcement Learning (RL) systems return an action no matter what situation they are confronted with, even for situations that differ entirely from those an agent has been trained for. In this work, we propose to formalize the changes in the environment in terms of the Markov Decision Process (MDP), resulting in a more formal framework when dealing with such problems.

: http://publica.fraunhofer.de/documents/N-640442.html