Apprentissage machine pour l'optimisation énergétique des réseaux cellulaires hétérogènes sans-fil : une approche bandit à bras multiples

Abstract : A machine learning algorithm is proposed in this paper to improve the energy efficiency (EE) of the heterogeneous cellular networks. The strategy aims at learning what is the optimal base stations (BS) configuration, i.e. ON or OFF, maximizing EE of the network. First, we formulate the traffic load variations as a Markov decision process (MDP), and present an algorithm from the upper confidence bound (UCB) class based on multi-armed bandit (MAB) theory to learn the best deployment of BS. Moreover, to cope with initial reward loss due to the variation of the load, a transfer learning (TL) framework for MAB, which benefits from the transferred knowledge observed in historical periods, is proposed. Extensive simulations demonstrate that the proposed solution can significantly increase EE of cellular networks.
Complete list of metadatas

Cited literature [7 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01720782
Contributor : Philippe Mary <>
Submitted on : Thursday, March 1, 2018 - 3:22:17 PM
Last modification on : Thursday, February 7, 2019 - 4:46:10 PM
Long-term archiving on : Wednesday, May 30, 2018 - 1:50:33 PM

File

BM_EE.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01720782, version 1

Citation

Navikkumar Modi, Philippe Mary, Christophe Moy. Apprentissage machine pour l'optimisation énergétique des réseaux cellulaires hétérogènes sans-fil : une approche bandit à bras multiples. GRETSI, Sep 2017, Juan-Les-Pins, France. ⟨hal-01720782⟩

Share

Metrics

Record views

1085

Files downloads

144