Cooperative comodule discovery for swarm-intelligent drone arrays

• Published in: Computer communications 2020-03, Vol.154, p.528-533
• Main Authors: Chuang, Hsin, Hou, Kuan-Lin, Rho, Seungmin, Chen, Bo-Wei
• Format: Article
• Language: English
• Subjects: Comodule; Cooperative mode; Graph signal processing; Nonnegative matrix factorization; Self-organizing map; Sparse network factorization; Swarm intelligence
• ISSN: 0140-3664; 1873-703X
• DOI: 10.1016/j.comcom.2020.02.077

This study presents a method for discovering comodules from a drone array. Herein, a comodule represents a frequent connection pattern among drones and their connected targets (e.g., mobile phones and end users) — Frequent connections among drone-to-drone (D2D) and drone-to-target (D2T) patterns. In the Internet of Drones (IoD), where sensing and communications become a major task, it is important to explore the cooperative components (i.e., subgraphs or comodules) of the network because data streams frequently flow through those components. Once a comodule is found, the control center can notify the drones along with their connected targets to process data in a cooperative mode. However, it is difficult to extract comodules since the topology of the IoD is fluid. Additionally, the targets connected to the IoD could be dynamic and mobile. To effectively discover comodules while considering mobile targets at the same time, this study proposes regularized self-organizing sparse network factorization. It jointly decomposes a connection network formed by a drone array and its connected targets into basis and coefficient matrices. Comodules can be explored by examining those matrices. Experiments were carried out on open Call Data Records and a drone array. The results show that the proposed method generated less divergence and higher tightness between the nodes of a comodule than the baseline did.