Description
Date depot: 21 novembre 2024
Titre: Adaptive AI-based multi-agent intelligent road infrastructure strategy for decentralized urban road traffic management
Directeur de thèse:
Jérôme HÄERRI (Eurecom)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini
Resumé: This thesis focuses on the analysis of the problematic to develop and optimize connected, reactive and fully decentralized traffic light infrastructure with the objective to improve traffic fluidity and reduced pollution.
If reactive smart traffic lights to optimize individual traffic lights have been well studied, the interaction with potentially competing signaling infrastructure with various complementary or opposite objectives remains challenging. In particular, AI has been studied to learn traffic patterns and adapt the traffic light policies.
However, if AI predictions can be altered by the decision of a different AI, approaches such as reinforcement learning might not even converge leading to both weak local and global decisions.
Accordingly, cooperative AI strategies would be required, which is currently an open research area, first from the very mechanisms to predict the prediction, as well as limiting the amount of data required to such cooperation.
Résumé dans une autre langue: Stratégie adaptative d'une infrastructure routière intelligente multi-agents pour une gestion décentralisée du trafic routier en milieu urbain.
Doctorant.e: Sauvage Michel