ENERGY EFFICIENT CLOCK SYNCHRONIZATION IN IOT USING REINFORCEMENT LEARNING

Loading...
Thumbnail Image

Date

2024-04-19

Authors

Nadirkhanova, Aizhuldyz
Assylbek, Damir

Journal Title

Journal ISSN

Volume Title

Publisher

Nazarbayev University School of Engineering and Digital Sciences

Abstract

This project addresses the vital challenge of achieving precise clock synchronization within the Internet of Things (IoT), a foundational element for the seamless and efficient operation of interconnected devices. Such synchronization is indispensable for critical IoT functions like coordinated actions, streamlined communication, and power management. The project introduces a novel approach leveraging Reinforcement Learning (RL), specifically the State-Action-Reward-State-Action (SARSA) algorithm. This method equips devices with the capability to autonomously learn and anticipate the timing of data transmissions, fostering self-synchronization without manual intervention or pre-programmed schedules. It's a significant shift from traditional manual adjustments of clock drift, accommodating the unique timing characteristics of each device's crystal oscillator. Small testbeds with ESP32 devices using the ESP-NOW protocol have validated the approach's adaptability to transmission timing variances, maintaining a high success rate in data receipt. Furthermore, the project continues to work on incorporating knowledge transfer techniques and Huffman coding to compress trained data, facilitating rapid convergence to optimal behavior and fostering an environment where devices benefit from shared learning experiences.

Description

Keywords

Machine Learning, Reinforcement Learning, Internet of Things, Synchronization, Wireless networks, Synchronization, Wireless networks, Type of access: Open Access

Citation

Assylbek, D., Nadirkhanova, A.(2024). Energy efficient Clock Synchronization in IoT Using Reinforcement Learning. Nazarbayev University School of Engineering and Digital Sciences