Formula omitted]: Energy-efficient intelligent edge–cloud collaboration for remote IoT services

Abstract

Existing centralized cloud-based solution is challenging to cope with the explosive growth of data generated from massive IoT devices to meet the requirements of critical services, especially for remote IoT services provided in remote/rural areas where the resource and energy capacity of local data processing infrastructure is limited. To solve this issue, we propose ELECT, an energy-efficient intelligent edge-cloud collaboration scheme that achieves satisfactory data processing performance. ELECT includes a platform that utilizes edge computing node as a core element that locates close to the IoT nodes for coordinating the data processing between the cloud and the IoT devices. Specifically, based on the importance of a node’s collected data in the overall data processing performance, a dynamic IoT node management algorithm is developed to manage each node’s active/inactive status to reduce energy consumption. Moreover, a deep Q-network (DQN)-based workflow scheduling algorithm that fully utilizes the data-centric device–edge–cloud continuum is introduced to reduce makespan and energy consumption for obtaining a compromising solution. For verification, we develop an experimental environment simulating structural health monitoring (SHM) services in remote areas. Extensive experiments verify the effectiveness of ELECT in terms of various service requirements, including makespan, energy consumption and communication cost.