Enhancing Energy Efficiency of Sensors and Communication Devices in Opportunistic Networks Through Human Mobility Interaction Prediction

TL;DR

This paper proposes using a random forest regressor to predict human mobility in opportunistic networks, improving energy efficiency and message routing performance.

Contribution

The study introduces a novel random forest regressor-based approach for mobility prediction that enhances energy efficiency in opportunistic networks.

Findings

The random forest regressor outperformed Gaussian process and existing methods in predicting mobility encounters.

Integration of mobility predictions reduced energy consumption by up to one-third in D2D and traditional networks.

The proposed approach addresses limitations of current mobility prediction models for opportunistic networks.

Abstract

The proliferation of smart devices such as sensors and communication devices has necessitated the development of networks that can adopt device-to-device communication for delay-tolerant data transfer and energy efficiency. Therefore, there is a need to develop opportunistic networks to enhance energy efficiency through improved data routing. A sensor device equipped with computing, communication, and mobility capabilities can opportunistically transfer data to another device, either as a direct recipient or as an intermediary forwarding data to a third device. Routing algorithms designed for such opportunistic networks aim to increase the probability of successful message transmission by leveraging area information derived from historical data to forecast potential encounters. However, accurately determining the precise locations of mobile devices remains highly challenging and necessitates a robust prediction mechanism to provide reliable insights into mobility encounters. In this study, we propose incorporating a random forest regressor (RFR) to predict the future location of mobile users, thereby enhancing message routing efficiency. The RFR utilizes mobility traces from diverse users and is equipped with sensors for computing and communication purposes. These predictions improve message routing performance and reduce energy and bandwidth resource utilization during routine data transmissions. To evaluate the proposed approach, we compared the predictive performance of the RFR against existing benchmark schemes, including the Gaussian process, using real-world mobility data traces. The mobility traces from the University of Southern California (USC) were employed to underpin the simulations. Our findings demonstrate that the RFR significantly outperformed both the Gaussian process and existing methods in predicting mobility encounters. Furthermore, the integration of mobility predictions into device-to-device (D2D) communication and traditional internet networks showed potential energy consumption reductions of up to one-third, highlighting the practical benefits of the proposed approach. The contribution of this research is that it highlights the limitations of existing mobility prediction models and develops new resource optimization and energy-efficient opportunistic networks that overcome these limitations.

Figures