Safe exploration for reinforcement learning
Autor(en): | Hans, A. Schneegaß, D. Schäfer, A.M. Udluft, S. |
Stichwörter: | Controlled system; Critical state; Safety constraint; Safety degree; Safety functions, Critical current density (superconductivity); Neural networks, Reinforcement learning | Erscheinungsdatum: | 2008 | Journal: | ESANN 2008 Proceedings, 16th European Symposium on Artificial Neural Networks - Advances in Computational Intelligence and Learning | Startseite: | 143 | Seitenende: | 148 | Zusammenfassung: | In this paper we define and address the problem of safe exploration in the context of reinforcement learning. Our notion of safety is concerned with states or transitions that can lead to damage and thus must be avoided. We introduce the concepts of a safety function for determining a state's safety degree and that of a backup policy that is able to lead the controlled system from a critical state back to a safe one. Moreover, we present a level-based exploration scheme that is able to generate a comprehensive base of observations while adhering safety constraints. We evaluate our approach on a simplified simulation of a gas turbine. |
Beschreibung: | Conference of 16th European Symposium on Artificial Neural Networks - Advances in Computational Intelligence and Learning, ESANN 2008 ; Conference Date: 23 April 2008 Through 25 April 2008; Conference Code:100606 |
ISBN: | 9782930307084 | Externe URL: | https://www.scopus.com/inward/record.uri?eid=2-s2.0-79956136559&partnerID=40&md5=d20e36d603bf5a6236a5e5659c8327c2 |
Zur Langanzeige
Seitenaufrufe
8
Letzte Woche
0
0
Letzter Monat
1
1
geprüft am 21.05.2024