Improving power consumption in IoT devices through neural network-based decision making

Efficient power management is a critical challenge in IoT-based fog computing environments due to limited energy resources, dynamic workloads, and heterogeneous devices. The existing method addresses this challenge by introducing a neural network–based decision-making framework for adaptive resource and power management. The system continuously monitors key operational parameters, including CPU utilization, power consumption, and the number of active containers on each fog node, to capture the real-time load and energy status of the infrastructure. These monitored metrics undergo a feature extraction process to generate meaningful representations of system states. The extracted features are used to train a Multilayer Perceptron (MLP) model, which learns the complex relationship between resource usage patterns and optimal migration decisions. Based on the learned behaviour, the neural network dynamically determines whether container migration is necessary and selects an appropriate destination, such as another fog node or the cloud, to balance workloads. This intelligent migration strategy aims to reduce excessive power consumption, prevent node overloading, and extend the battery lifetime of IoT devices. Compared to traditional static and rule-based approaches, the learning-driven framework offers improved adaptability to fluctuating workloads and changing network conditions. Overall, the method enhances energy efficiency and resource utilization in fog computing environments, making it suitable for scalable and sustainable IoT applications.

Keywords: Internet of Things; Energy Efficiency; Pareto Optimization; Neural Network Decision Making; Fog–Cloud Computing.