Spiking Neural Networks for Resource Allocation in UAV-Enabled Wireless Networks

TL;DR

This paper introduces SNN-based methods for optimizing user association in UAV-enabled wireless networks, demonstrating high accuracy and efficiency through simulations, and comparing centralized and distributed strategies for network management.

Contribution

The work presents a novel application of spiking neural networks for UE-BS association in heterogeneous UAV-enabled networks, including two distinct optimization strategies and their performance evaluation.

Findings

Bottom-up SNN model achieves over 90% accuracy.

Top-down SNN model maintains 80-100% accuracy.

Trade-off identified between optimality and feasibility in different scenarios.

Abstract

This work presents a new spiking neural network (SNN)-based approach for user equipment-base station (UE-BS) association in non-terrestrial networks (NTNs). With the introduction of UAV's in wireless networks, the system architecture becomes heterogeneous, resulting in the need for dynamic and efficient management to avoid congestion and sustain overall performance. The presented framework compares two SNN-based optimization strategies. Specifically, a top-down centralized approach with complete network visibility and a bottom-up distributed approach for individual network nodes. The SNN is based on leak integrate-and-fire neurons with temporal components, which can perform fast and efficient event-driven inference. Realistic ray-tracing simulations are conducted, which showcase that the bottom-up model attains over 90\% accuracy, while the top-down model maintains 80-100\% accuracy. Both approaches reveal a trade-off between individually optimal solutions and UE-BS association feasibility, thus revealing the effectiveness of both approaches depending on deployment scenarios.