Deep reinforcement learning-based dependent task offloading for Edge Intelligent Controllers

doi:10.1016/j.future.2026.108819

Deep reinforcement learning-based dependent task offloading for Edge Intelligent Controllers

2026-09-18

0

PRE

AI

DOI：10.1016/j.future.2026.108819

Original

Original request for help

Save

Abstract

En 中文

Edge Intelligent Controllers (EIC) for Industrial Internet of Things are widely used in smart manufacturing as core components of Industrial Edge Computing. In practical industrial control scenarios, a resource-constrained EIC cannot consistently satisfy the quality-of-service (QoS) requirements of computationally intensive tasks; consequently, selected tasks must be offloaded to an Edge Computing Server (ECS) for execution. Considering the dependencies among industrial application components, we model application tasks and their precedence relations as a directed acyclic graph (DAG). Then, we propose a sequence-to-sequence-based Deep Reinforcement Learning algorithm for Dependent Task Offloading (SDRL-DTO) to obtain an offloading strategy. The proposed model considers two optimization objectives: delay minimization and joint delay–energy utility maximization. Simulation results demonstrate that SDRL-DTO outperforms the evaluated non-exhaustive baselines across the considered settings; at a transmission rate of 20 Mbps, it achieves mean delay comparable to HEFT and Greedy while reducing EIC-side energy consumption by 30%.

Deep reinforcement learning-based dependent task offloading for Edge Intelligent Controllers

Abstract

Journal

Organization

Cited Papers