A REINFORCEMENT LEARNING-BASED FRAMEWORK FOR ADAPTIVE ENERGY MANAGEMENT AND COVERAGE CONTROL IN IOT NETWORKS
Mathematics for Applications · 16.12.2025 · 14 (2) · 131-144
DOI: 10.66838/ma.236 ↗Kurz gefasst
Fachartikel in der Zeitschrift Mathematics for Applications, online erschienen am 16.12.2025. Verfasst von Hemant Kumar und 2 weiteren Autor:innen.
Abstract (Originalsprache)
The Internet of Things (IoT) connects sensing devices, embedded controllers, gateways and cloud or edge platforms to support continuous monitoring and intelligent services. A large proportion of IoT deployments rely on battery-powered wireless sensor nodes whose energy resources are limited. Continuous sensing, idle listening, packet transmission, retransmission and redundant coverage can deplete batteries and shorten network lifetime. At the same time, placing too many nodes in sleep mode may create coverage holes, weaken connectivity and reduce the reliability of collected information. This paper proposes a reinforcement learning-based framework for jointly managing energy consumption and sensing coverage in IoT networks. The framework models network operation as a sequential decision problem in which an agent observes residual energy, node activity, local density, coverage contribution, traffic demand and link quality, and then selects actions such as node activation, sleep scheduling, sensing-rate adjustment, transmission-power control and relay selection. A Deep Q-Network (DQN) is proposed for learning action values in a state space that may be too large for a conventional tabular Q-learning approach. A multi-objective reward function is formulated to balance energy expenditure, coverage sufficiency, connectivity, packet delivery and delay. The framework also identifies a distributed multi-agent extension for large deployments in which local agents coordinate through limited neighbourhood information. Mathematical models are presented for residual energy, radio transmission, geometric coverage, network lifetime and constrained optimization. A reproducible simulation methodology is specified, including candidate baselines, experimental parameters and statistical evaluation procedures. Since no new simulation run is claimed in this paper, numerical improvements are not fabricated; the results section defines the experiments and the evidence required to validate the proposed approach. The framework provides a research foundation for sustainable monitoring in smart agriculture, environmental sensing, industrial IoT and smart infrastructure.
Zitieren
APA
Kumar, H., Singh, G., & Singh, S. (2025). A REINFORCEMENT LEARNING-BASED FRAMEWORK FOR ADAPTIVE ENERGY MANAGEMENT AND COVERAGE CONTROL IN IOT NETWORKS. Mathematics for Applications, 14(2), 131-144. https://doi.org/10.66838/ma.236
Vancouver
Kumar H, Singh G, Singh S. A REINFORCEMENT LEARNING-BASED FRAMEWORK FOR ADAPTIVE ENERGY MANAGEMENT AND COVERAGE CONTROL IN IOT NETWORKS. Mathematics for Applications. 2025;14(2):131-144. doi:10.66838/ma.236
BibTeX
@article{Kumar2025,
author = {Kumar, Hemant and Singh, Gurvinder and Singh, Shamsher},
title = {A REINFORCEMENT LEARNING-BASED FRAMEWORK FOR ADAPTIVE ENERGY MANAGEMENT AND COVERAGE CONTROL IN IOT NETWORKS},
journal = {Mathematics for Applications},
year = {2025},
volume = {14},
number = {2},
pages = {131--144},
doi = {10.66838/ma.236}
}
EUVIMED