Ergodicity through Mean Field Asymptotics – Application to Decentralized Control in Communication Networks
by
Alexandre Proutiere(KTH)
→
Europe/Stockholm
Description
Mean Field Asymptotics are classically used to approximate the dynamics of stochastic systems consisting of a large number of particles. Most often, the convergence towards the mean field limit when the system size grows large can be proved in transient regimes, i.e., over bounded time intervals. Sometimes, under strong assumptions, the convergence results can be extended to the stationary regime. In this talk, we try to push mean field asymptotics even further, and are interested in establishing a relationship between the ergodicity of the initial stochastic system and its global stability in the mean field regime. We exhibit conditions under which such relationship can be formally proved, and apply these results to derive asymptotically exact stability conditions for communication networks whose resources are shared using classical decentralized random multi-access protocols.