Carbon-aware resource allocation and task offloading in EH-assisted edge-cloud systems

With the development of the Internet of Things (IoT) and Information Communication Technology (ICT), a large number of computation-intensive terminal applications have emerged. Multi-access Edge Computing (MEC) is a promising technology to meet terminal devices’ computational and battery-capacity demands, yet it also generates significant energy consumption and carbon emissions. This paper investigates the dynamic task offloading problem in Energy Harvesting (EH)-assisted edge-cloud systems. We aim to minimize the overall carbon emissions of the system while ensuring long-term stability and satisfying multi-user, multi-server constraints by optimizing CPU frequencies, offloaded task sizes, and offloading decisions. Specifically, we formulate the problem as a multi-slot stochastic optimization problem, which is NP-hard. Employing a stochastic optimization framework, we transform and decompose the problem into multiple subproblems that can be solved efficiently. Considering the coupling between offloading decisions and resource allocation, we propose the Carbon-Aware Resource Allocation and Offloading (CARAE) algorithm to obtain an efficient carbon-aware task offloading strategy. A series of theoretical analyses and simulations are conducted to verify that the proposed CARAE algorithm effectively reduces overall carbon emissions while maintaining system performance.