Enhancing Energy Efficiency in Edge-Cloud Systems through Deep Reinforcement Learning-Based Task Scheduling

Intelligent decision-making frameworks driven by neural policy optimization are the sectors that this research is aimed at. The main objective is to increase the energy efficiency of edge-cloud systems by accomplishing task scheduling optimally. Moreover, the IoT devices grow in numbers; therefore, the edge-cloud systems become balance computational load challenging, and the energy efficiency of the systems becomes problematic. Scheduling tasks by traditional methods are quite inefficient in changing system states and task requests adapting dynamically. The article proposes a DRL-driven scheduling architecture that employs experience and even innovatively schedules policies from system states such as resource availability, task behaviors, and energy consumption. The DRL agent is trained by a reward function that seeks to maximize the energy efficiency without any loss of system throughput and minimum response time. With a large number of simulations and experiments on an edge and cloud server testbed, we show that our DRL-based scheduler is able to achieve a significant performance gain over classical heuristic techniques. It yields up to 15 to 35% in energy saving and a 15 to 20% decrease in the average task completion time when compared with the state-of-the-art scheduling methods. This framework is also highly scalable as it can handle numerous tasks and devices without deteriorating performance. The article discusses the advantages of DRL in devising flexible and energy-efficient task scheduling strategies for the edge-cloud systems as a promising future course of research and engineering implementation in IoT applications.