With the prevalence of real-time Internet of Things (IOT) applications, mobile edge computing (MEC) has emerged as a paradigm to ensure low-latency computation. Hierarchical MEC system, which deploy computation resources in a hetero-geneous way, is more applicable than traditional MEC system in scenarios with various types of offloading tasks. On the other hand, the boom of AI applications has led to an increasing energy consumption of MEC servers. Energy harvesting technologies effectively reduce the grid energy consumption by leveraging renewable energy sources. To this end, we propose a task offloading scheme in the scenario with hierarchical task offloading and green energy harvesting. Specifically, We formulate the offloading problem as minimizing the weighted sum of execution latency and energy consumption. To obtain the optimal offloading decisions, we propose a computing offloading algorithm based on deep reinforcement learning (DRL). Simulation results demonstrate that our proposed scheme efficiently reduces the sum cost compared with other approaches.
Learning-Based Computation Offloading in Hierarchical MEC System with Energy Harvesting
2024-06-24
1685430 byte
Conference paper
Electronic Resource
English