ENERGY-EFFICIENT SYNCHRONIZATION USING TRANSFER OF KNOWLEDGE

Abstract

The project tackles the problem of one-to-many Internet of Things (IoT) synchronization with the goal of energy efficiency optimization. It builds upon the previous work done by Assylbek et al. (2024) that implemented a Reinforcement Learning based solution to synchronize receiver end-devices with a central transmitter device that broadcasts beacons in a constant interval. Our solution adds a component of Transfer Learning to further enhance the energy efficiency by eliminating the need to learn from scratch. Devices that newly join the system can request and use the knowledge from the experienced devices, therefore decreasing their learning time and enabling them to start with closer to optimal behaviour. We refer to this process as Transfer of Knowledge (ToK). Since the energy consumption of IoT devices is tied to their sleep times and radio-on times, our goal and expectation in implementing ToK is to maximize the sleep time and minimize radio-on time on end-devices.