Self-Organizing Complex Networks with AI-Driven Adaptive Nodes for Optimized Connectivity and Energy Efficiency

TL;DR

The paper addresses the challenge of achieving guaranteed connectivity and robustness in distributed, energy-constrained networks. It introduces an AI-driven self-organizing framework in which each node runs MLPs trained on a Hamiltonian-derived dataset to autonomously tune transmission power and manage link formation. The key contributions include a dual-MLP node architecture, a physics-informed training dataset, and extensive simulations showing stable full connectivity, robustness to failures, and energy efficiency in both static and mobile 2D/3D networks. The work demonstrates the practical potential of physics-guided machine learning for scalable, resilient distributed systems in IoT, WSNs, and autonomous networks.

Abstract

High connectivity and robustness are critical requirements in distributed networks, as they ensure resilience, efficient communication, and adaptability in dynamic environments. Additionally, optimizing energy consumption is also paramount for ensuring sustainability of networks composed of energy-constrained devices and prolonging their operational lifespan. In this study, we introduce an Artificial Intelligence (AI)-enhanced self-organizing network model, where each adaptive node autonomously adjusts its transmission power to optimize network connectivity and redundancy while lowering energy consumption. Building on our previous Hamiltonian-based methodology, which is designed to lead networks toward globally optimized states of complete connectivity and minimal energy usage, this research integrates a Multi-Layer Perceptron (MLP)-based decision-making model at each node. By leveraging a dataset from the Hamiltonian approach, each node independently learns and adapts its transmission power in response to local conditions, resulting in emergent global behaviors marked by high connectivity and resilience against structural disruptions. This distributed, MLP-driven adaptability allows nodes to make context-aware power adjustments autonomously, enabling the network to maintain its optimized state over time. Simulation results show that the proposed AI-driven adaptive nodes collectively achieve stable complete connectivity, significant robustness, and optimized energy usage under various conditions, including static and mobile network scenarios. This work contributes to the growing field of self-organizing networks by illustrating the potential of AI to enhance complex network design, supporting the development of scalable, resilient, and energy-efficient distributed systems across diverse applications.

Figures (9)

• Figure 1: Illustration of adaptive network nodes and their connections based on Euclidean distance and transmission range, focusing on a specific spatial configuration within the network. Nodes are homogeneous in capabilities, and links are established between any two nodes if they are within each other’s transmission range seyyedi2023energy.
• Figure 2: Snapshots and connectivity matrices of the complex network of adaptive nodes at steps 0, 200, 500, and 2500. (a) Static network snapshots showing the evolution of node connectivity and the merging of components over time until reaching full connectivity, with nodes displayed in different colors to distinguish components. (b) Connectivity matrices for static network snapshots, where yellow cells indicate connections between node pairs. (c) Mobile network snapshots, similar to (a), showing node connectivity with movement introduced over time. (d) Connectivity matrices for mobile network snapshots, reflecting network progression with node movement. The networks are spatial, with x and y coordinates considered in the snapshots (a) and (c). Connectivity matrices (b) and (d) are symmetric, reflecting undirected, bidirectional links.
• Figure 3: This figure presents the evolution of key network metrics over 10,000 steps, averaged over 10 different initial configurations for each network type. The error bars reflect the variability among these simulations, capturing the robustness of the results despite differences in initial node configurations. Subfigures (a) and (b) depict metrics for the complex network of static adaptive nodes, while (c) and (d) show equivalent metrics for mobile adaptive nodes, all at a node distribution density of 0.05. The metrics tracked include connectivity, average range, energy usage, link count ratio, coverage, global clustering, local clustering, and average degree. The data reveals how both static and mobile networks reach stable connectivity and other desired states, with mobile networks adapting under the challenge of random movement dynamics.
• Figure 4: This figure illustrates the evolution of network coverage over time for two types of adaptive networks: static (a) and mobile (b). The snapshots are taken at steps 0, 200, 500, and 2500, demonstrating how each network adapts its coverage area as nodes adjust their transmission ranges. The blue circles represent the transmission range of individual nodes, while the red dashed outline shows the boundary of the total covered area. In both cases, the networks begin with partial coverage and progressively expand their reach, achieving full coverage.
• Figure 5: Degree distribution histograms of the adaptive complex network, comparing static and mobile node scenarios at different time steps (0, 200, 500, and 2500). The histograms show the evolution of node connectivity (node degree) as the adaptive algorithm increases network connectivity while maintaining energy efficiency. Subfigure (a) represents the static network, while subfigure (b) represents the mobile network, illustrating the impact of node mobility on the degree distribution over time.