Authors:
Prasanna Ranjith Christodoss, Rahul Chauhan, Bhargav Gadi, Abduljaleel Al-Janabi, Dinka Lale, Ramadas Kandath

Addresses:
Department of Computing, Mathematics and Physics, Messiah University, Mechanicsburg, Pennsylvania, United States of America. Department of Fitch Ratings, The Fitch Group, New York, United States of America. Department of Computer Science and Engineering, SRM Institute of Science and Technology, Ramapuram, Chennai, Tamil Nadu, India. Department of Environmental Engineering, University of Babylon, Al-Hilla, Babylon, Iraq. Faculty of Electrical Engineering and Applied Computing, University of Dubrovnik, Dubrovnik, Dubrovnik-Neretva County, Croatia. Department of FSM Adoption and Services, SAP, Dallas, Texas, United States of America.

Abstract:

Driving an electric vehicle in uncertain situations requires trusting your battery range and finding an available charger. NASK-ML confronts these obstacles. Researchers combined high-frequency vehicle data with privacy-preserving network techniques (BLAKE3 IP masking). Researchers added machine-learning modules to anticipate short-term range, rank reachable chargers, automate charge-versus-tow decisions, and detect network irregularities. Our approach treats vehicle telemetry as a temporal signal, not a formula. The solution gives ordinary drivers and fleet operators the most accurate recommendations by integrating live data with past station data. Of course, reality is messy. Researchers had to account for poor internet connectivity, edge-device compute constraints, and privacy demands. NASK-ML's hybrid architecture handles this friction. CNNs, MLPs, LSTMs, and Transformers were coupled by attention mechanisms to encode spatial and temporal data. Researchers kept ranking models lightweight to sort candidates on the device. So raw telemetry never leaves the vehicle, researchers added an optional federated training pathway. Researchers used SHAP and time-series saliency maps to show operators why the system made a safety judgement. Researchers found clear findings in our simulated NASK logs. Our hybrid models outperformed typical heuristics, and federated training showed that user privacy may be protected without compromising system value.

Keywords: Electric Vehicle; High-Frequency Vehicle Data; Machine-Learning Modules; BLAKE3 IP Masking; NASK-ML; Historical Station Information; Short-Term Range.

Received on: 01/07/2025, Revised on: 10/09/2025, Accepted on: 03/11/2025, Published on: 09/06/2026

DOI: 10.69888/FTSIS.2026.000773

FMDB Transactions on Sustainable Intelligence and Security, 2026 Vol. 1 No. 2, Pages: 102-117

• Views : 40
• Downloads : 9

Download PDF