Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
August 1, 2026TechnologiesOpen Access

Robust Offline Multi-Agent Reinforcement Learning for Latency-Aware SDN Path Control in 6G-Oriented Network Softwarization

View Full Paper
Ask AI
Bookmark
Share

Authors

АКАбзал КызыркановYNYedil NurakhovZOZhenis Otarbay

Discussion

Loading...

Member takes

Overview

Randomized trial evaluates offline learning for latency-aware routing in software-defined networks, suggesting a competitive low-overhead option.

Key Points

  • The aim is to assess offline multi-agent reinforcement learning for effective latency-aware path control in 6G-oriented networks.
  • Utilized offline multi-agent deep deterministic policy gradient (MADDPG) with behavior-adjusted rewards.
  • Compared nine policies across various topologies (fat-tree, mesh-grid, WAN-corridors) with traffic pairs modeled as agents.
  • Performed analyses on flow-completion, latency, congestion, and controller overhead.
  • The utilization-aware path heuristic achieved the highest overall reward ranking.
  • MADDPG outperformed other policies on fat-tree topology and showed strong performance on mesh-grid and WAN-corridors.
  • Behavior adjustment effectiveness was found to be topology-dependent.

Cite This Study

Кызырканов et al. (2026) studied this question.

synapsesocial.com/papers/6a6da778e258b358b3c6c8a0https://doi.org/10.3390/technologies14080468
View Full Paper
Ask AI
Bookmark
Share