@article{Raut2026, 
author = {Nitin B. Raut and Shanmuga Sundaram Thangavelu},
title = {Optimized clustering and deep deterministic policy gradient mechanism for energy efficient sleep scheduling in green IoT},
year = {2026},
journal = {Intelligent and Converged Networks},
volume = {7},
number = {3},
pages = {327-346},
keywords = {green Internet of Things (IoT), Horse Herd Optimization Algorithm (HOA), Affinity Propagation (AP), Deep Deterministic Policy Gradient Algorithm (DDPGA), energy efficient clustering},
url = {https://www.sciopen.com/article/10.23919/ICN.2025.0027},
doi = {10.23919/ICN.2025.0027},
abstract = {Old industrial system is projected to become more intelligent and efficient with the help of the growing Internet of things (IoT) concept. Emerging as a new paradigm, the green IoT envisions the idea of connecting various equipment and lowering energy use. Allowing some sensor nodes to remain awake during a certain period of time while permitting others to sleep is a common technique for maintaining the sensor nodes energy. Using a sleep scheduling algorithm to dynamically arrange the active/sleep cycles (also known as duty cycles) of sensors is one typical method for extending the lifetime of a Wireless Sensor Networks (WSN). However, the existing algorithms are unable to guarantee a uniform consumption of network energy. So, in this proposed approach, an energy efficient optimized clustering and reinforcement learning based sleep scheduling algorithm is developed. The deployed sensor node initially goes through the clustering phase using the Horse Herd Optimization Algorithm (HOA) and Affinity Propagation (AP) algorithms. HOA is employed for the number of exemplar selection in AP. Following that, the Deep Reinforcement Learning (DRL) algorithm is used to adopt duty cycling in the second phase. This work separates the time period into many time slots that are allotted to each sensor node. To avoid data conflicts during transmission, every sensor node has a separate slot. In every slot, nodes adaptively change their mode such as sleep or wakeup using a reinforcement learning based on the Deep Deterministic Policy Gradient Algorithm (DDPGA). Then sleep cycle is computed by using Fuzzy Interference System based on some parameters. This proposed method achieves 8.56%, 1.26 ms of transmission delay, and 13.7% packet loss of packet delivery ratio. This methodology efficiently minimizes the energy consumed by sensor nodes in a green IoT environment.}
}