Projet ANR PROPAGATE

Control of propagation processes on very large networks

The overarching objective of the PROPAGATE project is to develop a unified framework for de- signing efficient control mechanisms for stochastic propagation processes over massive-scale networks. The core novelty of our approach lies in shifting from conventional node-based policies to interaction- level control (graph-based). To achieve this, we will first develop new models that approximate network behavior using graphs that evolve in response to the applied control. A primary challenge here is en- suring model validity despite the tight coupling between propagation and graph dynamics—a scenario where classical approximations often fail, as demonstrated in [Kra+23]. The second phase of the project focuses on the scalability and complexity of these control laws. For finite-dimensional systems, we will design dynamic controllers to steer the propagation process while optimizing control costs. For systems where the number of nodes tends toward infinity, we will optimize the graphon kernel to maintain analytical tractability. Finally, we will employ game-theoretic models to derive optimal strategies in multi-agent environments, specifically addressing scenarios where individual nodes act selfishly to determine their own local control laws.

Liste des partenaires :

Responsable Scientifique pour le LIA : Yezekael Hayel (Equipe CORNET )

Date Début : 2026-10-01 Date Fin : 2030-10-01