Fast and Adaptive Task Management in MEC: A Deep Learning Approach Using Pointer Networks

TL;DR

This paper introduces a deep learning-based Pointer Network architecture for task scheduling in Mobile Edge Computing, achieving faster inference and better adaptability than traditional methods in dynamic, real-time environments.

Contribution

The paper presents a novel Pointer Network model trained on synthetic data for scalable, adaptive task scheduling in MEC, outperforming conventional algorithms in speed and flexibility.

Findings

Lower drop ratios and waiting times compared to baselines

Inference times under 2 seconds across task sizes

Strong generalization and adaptability to real-time changes

Abstract

Task offloading and scheduling in Mobile Edge Computing (MEC) are vital for meeting the low-latency demands of modern IoT and dynamic task scheduling scenarios. MEC reduces the processing burden on resource-constrained devices by enabling task execution at nearby edge servers. However, efficient task scheduling remains a challenge in dynamic, time-sensitive environments. Conventional methods -- such as heuristic algorithms and mixed-integer programming -- suffer from high computational overhead, limiting their real-time applicability. Existing deep learning (DL) approaches offer faster inference but often lack scalability and adaptability to dynamic workloads. To address these issues, we propose a Pointer Network-based architecture for task scheduling in dynamic edge computing scenarios. Our model is trained on a generated synthetic dataset using genetic algorithms to determine the optimal task ordering. Experimental results show that our model achieves lower drop ratios and waiting times than baseline methods, and a soft sequence accuracy of up to 89.2%. Our model consistently achieves inference times under 2 seconds across all evaluated task counts, whereas the integer and binary programming approaches require approximately up to 18 seconds and 90 seconds, respectively. It also shows strong generalization across varying scenarios, and adaptability to real-time changes, offering a scalable and efficient solution for edge-based task management.