5G standardization has envisioned mmWave communications as a promising direction to expand the capacity of current mobile radio networks. However, communications at high frequency are characterized by extremely harsh propagation conditions, thus requiring a high base station deployment density. To solve this issue, from both technical and economic perspective, 3GPP has proposed mmWave access networks based on an Integrated Access and Backhaul (IAB) multi-hop architecture.IAB networks require fine-tuning of the available resources in a complex setting, due to directional transmissions, device heterogeneity, and harsh propagation conditions. The latter, in particular, characterize the operations of such networks, resulting in links with very different levels of availability. For this reason, traditional optimization techniques do not provide the best performance in these conditions. We believe, instead, Reinforcement Learning (RL) techniques can implicitly consider the dynamics of the network links and learn the best resource allocation strategy in networks with intermittent links. In this paper, we propose an RL-based resource allocation approach that shows the advantages of these techniques in dynamic environmental conditions.

RL-based Resource Allocation in mmWave 5G IAB Networks

Zhang B.;Devoti F.;Filippini I.
2020-01-01

Abstract

5G standardization has envisioned mmWave communications as a promising direction to expand the capacity of current mobile radio networks. However, communications at high frequency are characterized by extremely harsh propagation conditions, thus requiring a high base station deployment density. To solve this issue, from both technical and economic perspective, 3GPP has proposed mmWave access networks based on an Integrated Access and Backhaul (IAB) multi-hop architecture.IAB networks require fine-tuning of the available resources in a complex setting, due to directional transmissions, device heterogeneity, and harsh propagation conditions. The latter, in particular, characterize the operations of such networks, resulting in links with very different levels of availability. For this reason, traditional optimization techniques do not provide the best performance in these conditions. We believe, instead, Reinforcement Learning (RL) techniques can implicitly consider the dynamics of the network links and learn the best resource allocation strategy in networks with intermittent links. In this paper, we propose an RL-based resource allocation approach that shows the advantages of these techniques in dynamic environmental conditions.
2020
Proceedings of 2020 Mediterranean Communication and Computer Networking Conference, MedComNet 2020
978-1-7281-6248-5
File in questo prodotto:
File Dimensione Formato  
2020_MedComNet_RL-IAB.pdf

accesso aperto

: Pre-Print (o Pre-Refereeing)
Dimensione 960.05 kB
Formato Adobe PDF
960.05 kB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11311/1170593
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 8
  • ???jsp.display-item.citation.isi??? 7
social impact