Toward a self-learning governance loop for competitive multi-attribute MAS

PDF
Toward_a_Self_Learning_Governance_Loop_for_Competitive_Multi_Attribute_MAS.pdf - Angenommene Version
Download (294kB)

URL:	https://madoc.bib.uni-mannheim.de/59649
Weitere URL:	https://dl.acm.org/doi/abs/10.5555/3463952.3464179
URN:	urn:nbn:de:bsz:180-madoc-596492
Dokumenttyp:	Konferenzveröffentlichung
Erscheinungsjahr:	2021
Buchtitel:	AAMAS '21: Proceedings of the 20th International Conference on Autonomous Agents and MultiAgent Systems : Richland, SC, Virtual Event United Kingdom, May, 2021
Seitenbereich:	1619-1621
Veranstaltungstitel:	AAMAS 2021
Veranstaltungsort:	Online
Veranstaltungsdatum:	03.-07.05.2021
Herausgeber:	Endriss, Ulle ; Nowé, Anne ; Dignum, Frank ; Lomuscio, Alessio
Ort der Veröffentlichung:	Richland, SC
Verlag:	International Foundation for Autonomous Agents and Multiagent Systems
ISBN:	978-1-4503-8307-3
Verwandte URLs:	https://generic.wordpress.soton.ac.uk/aa... http://www.ifaamas.org/Proceedings/aamas...
Sprache der Veröffentlichung:	Englisch
Einrichtung:	Außerfakultäre Einrichtungen > Institut für Enterprise Systems (InES)
Fachgebiet:	004 Informatik
Freie Schlagwörter (Englisch):	Restriction , Competition , Governance, Multi-Agent System
Abstract:	Competitive Multi-Agent Systems (MAS) are inherently hard to control due to agent autonomy and strategic behavior, which is particularly problematic when there are system-level objectives to be achieved or specific environmental states to be avoided.Existing methods mostly assume specific knowledge about agent preferences, utilities and strategies, neglecting the fact that actions are not always directly linked to genuine agent preferences, but can also reflect anticipated competitor behavior, be a concession to a superior adversary or simply be intended to mislead other agents. This assumption both reduces applicability to real-world systems and opens room for manipulation.We therefore propose a new governance approach for Multi-Attribute MAS which relies exclusively on publicly observable actions and transitions, and uses the acquired knowledge to purposefully restrict action spaces, thereby achieving the system's objectives while preserving a high level of autonomy for the agents.