Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning Power Control Protocol for In-Factory 6G Subnetworks

Uyoata E. Uyoata, Gilberto Berardinelli, Ramoni Adeogun

TL;DR

This work tackles interference management and signaling overhead in dense In-Factory 6G subnetworks by casting power-control as a multi-agent reinforcement learning problem within a POMDP framework. It employs MAPPO, with the central controller acting as an expert, to learn signaling and power-control policies for autonomous APs. Results show the MAPPO-based protocol greatly reduces signaling overhead—by about a factor of 8 relative to a Genie baseline—while achieving a near-optimal success rate close to the Genie solution, demonstrating practical viability in dynamic factory environments. The approach offers scalable, autonomous resource management for tight latency and reliability requirements in next-generation industrial networks.

Abstract

In-X Subnetworks are envisioned to meet the stringent demands of short-range communication in diverse 6G use cases. In the context of In-Factory scenarios, effective power control is critical to mitigating the impact of interference resulting from potentially high subnetwork density. Existing approaches to power control in this domain have predominantly emphasized the data plane, often overlooking the impact of signaling overhead. Furthermore, prior work has typically adopted a network-centric perspective, relying on the assumption of complete and up-to-date channel state information (CSI) being readily available at the central controller. This paper introduces a novel multi-agent reinforcement learning (MARL) framework designed to enable access points to autonomously learn both signaling and power control protocols in an In-Factory Subnetwork environment. By formulating the problem as a partially observable Markov decision process (POMDP) and leveraging multi-agent proximal policy optimization (MAPPO), the proposed approach achieves significant advantages. The simulation results demonstrate that the learning-based method reduces signaling overhead by a factor of 8 while maintaining a buffer flush rate that lags the ideal "Genie" approach by only 5%.

Learning Power Control Protocol for In-Factory 6G Subnetworks

TL;DR

This work tackles interference management and signaling overhead in dense In-Factory 6G subnetworks by casting power-control as a multi-agent reinforcement learning problem within a POMDP framework. It employs MAPPO, with the central controller acting as an expert, to learn signaling and power-control policies for autonomous APs. Results show the MAPPO-based protocol greatly reduces signaling overhead—by about a factor of 8 relative to a Genie baseline—while achieving a near-optimal success rate close to the Genie solution, demonstrating practical viability in dynamic factory environments. The approach offers scalable, autonomous resource management for tight latency and reliability requirements in next-generation industrial networks.

Abstract

In-X Subnetworks are envisioned to meet the stringent demands of short-range communication in diverse 6G use cases. In the context of In-Factory scenarios, effective power control is critical to mitigating the impact of interference resulting from potentially high subnetwork density. Existing approaches to power control in this domain have predominantly emphasized the data plane, often overlooking the impact of signaling overhead. Furthermore, prior work has typically adopted a network-centric perspective, relying on the assumption of complete and up-to-date channel state information (CSI) being readily available at the central controller. This paper introduces a novel multi-agent reinforcement learning (MARL) framework designed to enable access points to autonomously learn both signaling and power control protocols in an In-Factory Subnetwork environment. By formulating the problem as a partially observable Markov decision process (POMDP) and leveraging multi-agent proximal policy optimization (MAPPO), the proposed approach achieves significant advantages. The simulation results demonstrate that the learning-based method reduces signaling overhead by a factor of 8 while maintaining a buffer flush rate that lags the ideal "Genie" approach by only 5%.
Paper Structure (21 sections, 6 equations, 4 figures, 2 tables, 1 algorithm)

This paper contains 21 sections, 6 equations, 4 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. System Model
  3. RL based learned protocol for power control
  4. Agent Observation Space
  5. Agent Action Space
  6. Agent State Space
  7. Power Allocation Algorithm
  8. Agent Reward
  9. Performance Evaluation
  10. MAPPO Architecture
  11. Benchmarks
  12. Fixed Action
  13. Random Action
  14. Interference-Aware
  15. CSI Based Interference-Aware
  16. ...and 6 more sections

Figures (4)

  • Figure 1: A network of M subnetworks showing a CC, access points (APs) and sensor/actuator unit. Signaling link are indicated by green arrows, measurements are indicated by the the blue arrow, control signals are indicated by the red arrows. Each AP has a maximum of P packets in its buffer.
  • Figure 2: CDF of success rate of the considered approaches
  • Figure 3: Signaling overhead of the considered approaches
  • Figure 4: Action Selection Probability