Research Article
AIM-SGNet: An AI Agent-Driven Fuzzy MCDA Path Selection Method for Smart Grid IoT N2N Networks
@INPROCEEDINGS{10.1007/978-3-032-43712-9_5, author={Mei-Yu Wu and Chih-Kun Ke}, title={AIM-SGNet: An AI Agent-Driven Fuzzy MCDA Path Selection Method for Smart Grid IoT N2N Networks}, proceedings={Smart Grid and Internet of Things. 10th EAI International Conference, SGIoT 2026, Taichung, Taiwan, November 21--22, 2026, Proceedings}, proceedings_a={SGIOT}, year={2026}, month={11}, keywords={Smart Grid IoT N2N Networks AI Agent Fuzzy MCDA Network Path Selection}, doi={10.1007/978-3-032-43712-9_5} }- Mei-Yu Wu
Chih-Kun Ke
Year: 2026
AIM-SGNet: An AI Agent-Driven Fuzzy MCDA Path Selection Method for Smart Grid IoT N2N Networks
SGIOT
Springer
DOI: 10.1007/978-3-032-43712-9_5
Abstract
Smart Grid IoT applications generate heterogeneous communication traffic among control centers, substations, smart meters, distributed energy resources, edge servers, and sensor gateways. These traffic types have different network requirements. Fault detection and self-healing control require low latency and reliable delivery; smart meter reporting is more sensitive to cost and energy consumption, and distributed energy monitoring must balance bandwidth, data integrity, and security. Therefore, path selection based only on hop count, latency, or another single metric is insufficient for Smart Grid IoT N2N communication. This paper proposes Fuzzy AIM-SGNet, an AI-agent-assisted path selection method for Smart Grid IoT N2N networks. The agent observes network telemetry and grid task types, converts uncertain network conditions and task preferences into fuzzy linguistic variables, applies Fuzzy AHP to derive task-aware criteria weights, and uses Fuzzy TOPSIS to rank candidate paths. The method is evaluated using a NetworkX-based Smart Grid IoT topology and compared with shortest path, lowest latency, highest security, static weighted sum, and crisp TOPSIS baselines. The evaluation examines SLA violation rate, task satisfaction, recovery behavior, and decision interpretability.