Skip navigation
Please use this identifier to cite or link to this item: https://repository.esi-sba.dz/jspui/handle/123456789/1014
Full metadata record
DC FieldValueLanguage
dc.contributor.authorTAHIR, ABderrahmane El MEhdi-
dc.date.accessioned2026-10-11T07:35:20Z-
dc.date.available2026-10-11T07:35:20Z-
dc.date.issued2026-
dc.identifier.urihttps://repository.esi-sba.dz/jspui/handle/123456789/1014-
dc.descriptionSupervisor : Dr. KHALDI Belkacemen_US
dc.description.abstractMulti-agent reinforcement learning (MARL) has proven to be a powerful tool in multi-agent decision-making problems, and it has been employed in several contexts such as robotics, autonomous transportation, and warehouse management. However, MARL still faces many challenges, with the most significant being non-stationarity. This arises when multiple agents learn simultaneously, continuously altering the environment dynamics, and this problem is exacerbated when the learning environment is already non-stationary. Evolutionary game theory (EGT) on the other hand, is a branch of game theory that studies how successful strategies evolve in populations. This thesis investigates how EGT provides a complementary perspective by modeling the collective evolution of strategies and guiding convergence toward equilibria. In particular, we explore how learning dynamics from EGT could be used in MARL, and how their properties affect non-stationarity.*** L’apprentissage par renforcement multi-agents (MARL) s’est r´ev´el´e ˆetre un outil puissant pour les probl`emes de prise de d´ecision impliquant plusieurs agents, et il a ´et´e utilis´e dans divers contextes tels que la robotique, le transport autonome et la gestion d’entrepˆots. Cependant, le MARL fait encore face `a de nombreux d´efis, dont le plus important est la non-stationnarit´e. Celle-ci apparaˆıt lorsque plusieurs agents apprennent simultan´ement, modifiant en permanence la dynamique de l’environnement, et ce probl`eme est accentu´e lorsque l’environnement d’apprentissage est d´ej`a non stationnaire. La th´eorie des jeux ´evolutionnaire (EGT), quant `a elle, est une branche de la th´eorie des jeux qui ´etudie comment des strat´egies efficaces ´evoluent au sein de populations. Cette th`ese examine comment l’EGT offre une perspective compl´ementaire en mod´elisant l’´evolution collective des strat´egies et en guidant la convergence vers des ´equilibres. En particulier, nous explorons comment les dynamiques d’apprentissage issues de l’EGT peuvent ˆetre utilis´ees en MARL, ainsi que l’impact de leurs propri´et´es sur la non-stationnarit´e.en_US
dc.language.isoenen_US
dc.subjectMulti-Agent Reinforcement Learningen_US
dc.subjectEvolutionary Game Theoryen_US
dc.subjectNon-Stationary Environmentsen_US
dc.subjectEvolutionary Dynamicsen_US
dc.titleIntegrating Evolutionary Game Theory in Multi-Agent Reinforcement Learning for Non-Stationary Environmentsen_US
dc.typeThesisen_US
Appears in Collections:Master

Files in This Item:
File Description SizeFormat 
Masters_Thesis-1-1.pdf98,86 kBAdobe PDFView/Open
Show simple item record


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