Enhancing Sustainable Edge Computing Offloading via Renewable Prediction for Energy Harvesting

Abstract

The integration of Edge Computing (EC) and Energy Harvesting (EH) technologies has facilitated the growth of the Internet of Things (IoT), allowing for the interconnectivity of a wide range of devices. The integration of this technology has not only enhanced energy sustainability but also significantly extended the battery life of these devices. Adopting Renewable Energy (RE) sources has become more widespread in energy systems as a strategy to reduce carbon emissions. Low energy consumption and constrained battery capacity for IoT devices are concerns related to offloading. The unpredictability of RE quality makes it difficult for edge servers to maintain high quality of service in EH-EC systems, which impedes effective energy conservation for IoT. To solve an optimization problem, RE Predictions with a Deep Reinforcement Learning algorithm named (REP-DRL) are proposed. Accurately, REP-DRL used the actor-critic technique to identify the best approach for predicting RE and optimal offloading decisions. The approach improves IoT device processing and expands the system state to offload experiences per time slot. To store excess energy during periods of abundance and use it during times of higher demand, the service offloading process is modelled based on the predicted amount of RE, to find the best service offloading technique and improve energy sustainability for IoT. By determining the most efficient service offloading approach using the predicted RE amount, this solution increases the energy sustainability of the IoT ecosystem. Finally, the simulation results show that the REP-DRL system utilizes local computing to conserve power when both the battery level and projected EH are low, showcasing its capacity to adapt to varying operating conditions and optimize the utilization of resources.

Keywords

• Edge computing
• energy harvesting
• IoT
• RE
• DRL