Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

EdgeAgentX: A Novel Framework for Agentic AI at the Edge in Military Communication Networks

Abir Ray

TL;DR

EdgeAgentX tackles autonomous edge AI in military networks with denied or disrupted connectivity by integrating federated learning, MADDPG-based multi-agent reinforcement learning, and adversarial defenses into a cohesive three-layer framework. The approach enables scalable, cooperative policy learning across edge devices while preserving data privacy and security, and it demonstrates resilience to adversarial disruptions. In simulation, EdgeAgentX achieves lower end-to-end latency and higher throughput than decoupled baselines and converges faster, approaching the performance of a centralized controller without centralized data access. The work points to practical potential for field deployments in mission-critical edge networks and outlines directions for real-world testing, scalability improvements, and enhanced adversarial resilience.

Abstract

This paper introduces EdgeAgentX, a novel framework integrating federated learning (FL), multi-agent reinforcement learning (MARL), and adversarial defense mechanisms, tailored for military communication networks. EdgeAgentX significantly improves autonomous decision-making, reduces latency, enhances throughput, and robustly withstands adversarial disruptions, as evidenced by comprehensive simulations.

EdgeAgentX: A Novel Framework for Agentic AI at the Edge in Military Communication Networks

TL;DR

EdgeAgentX tackles autonomous edge AI in military networks with denied or disrupted connectivity by integrating federated learning, MADDPG-based multi-agent reinforcement learning, and adversarial defenses into a cohesive three-layer framework. The approach enables scalable, cooperative policy learning across edge devices while preserving data privacy and security, and it demonstrates resilience to adversarial disruptions. In simulation, EdgeAgentX achieves lower end-to-end latency and higher throughput than decoupled baselines and converges faster, approaching the performance of a centralized controller without centralized data access. The work points to practical potential for field deployments in mission-critical edge networks and outlines directions for real-world testing, scalability improvements, and enhanced adversarial resilience.

Abstract

This paper introduces EdgeAgentX, a novel framework integrating federated learning (FL), multi-agent reinforcement learning (MARL), and adversarial defense mechanisms, tailored for military communication networks. EdgeAgentX significantly improves autonomous decision-making, reduces latency, enhances throughput, and robustly withstands adversarial disruptions, as evidenced by comprehensive simulations.

Paper Structure

This paper contains 13 sections, 2 equations, 2 figures.

Table of Contents

  1. Introduction
  2. Proposed Framework (EdgeAgentX)
  3. Federated Learning Layer (Global Coordination)
  4. Multi-Agent RL Layer (Distributed Edge Intelligence)
  5. Adversarial Defense Layer (Security and Robustness)
  6. System Model and Algorithm
  7. Learning Algorithm
  8. Experimental Evaluation
  9. Evaluation Scenario
  10. Metrics
  11. Baselines
  12. Results
  13. Conclusion

Figures (2)

  • Figure 1: Conceptual architecture of EdgeAgentX showing the three-layer design -- a federated learning coordination layer (global), a multi-agent reinforcement learning layer (distributed edge intelligence), and an adversarial defense layer (security and robustness).
  • Figure 2: Learning curves comparing the convergence of EdgeAgentX with baseline approaches over training episodes. EdgeAgentX reaches a higher global reward faster than independent RL and federated RL without MARL, approaching the performance of a centralized RL agent much more rapidly.