Machine Learning Driven Latency Optimization for Internet of Things Applications in Edge Computing

Aug 5, 2026

Upload by Himbo

Emerging Internet of Things (IoT) applications demand lower execution time and response time, while many IoT devices lack sufficient onboard computing capability. Edge computing addresses this by moving computation closer to end devices via computation offloading. However, naive offloading that ignores resource demands, inter-application dependencies, and edge availability can cause execution delays and performance degradation. Edge-IoT introduces a machine learning-enabled orchestration framework and a bin-packing optimization model to achieve resource-aware offloading and reduced average latency.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages13

Reading Time33 min

AccessFree to download

Introduction

Edge computing and IoT offloading challenge

Motivation and problem statement

Abstracted approach

Edge-IoT orchestration framework

Resource-aware offloading and bin-packing optimization

Experiments and results

Effectiveness and resource efficiency

Special Topic Machine Learning Driven Latency Optimization for Internet of Things Applications in Edge Computing Uchechukwu AWADA, ZHANG Jiankang, CHEN Sheng, LI Shuangzhi, YANG Shouyi Machine Learning Driven Latency Optimization for Internet of Things Applications in Edge Computing Uchechukwu AWADA1 , ZHANG Jiankang2 , CHEN Sheng3 ,4 , LI Shuangzhi1 , YANG Shouyi1 (1 . Zhengzhou University , Zhengzhou 450001 , China； 2. Bournemouth University , Poole BH12 5BB , UK； 3. University of Southampton , Southampton SO17 1 BJ, UK； 4. Ocean University of China , Qingdao 266100 , China) DOI: 10. 12142/ZTECOM.202302007 [https://kns.cnki. net/kcms/detail/34.1294.TN.20230516.1317.002.html](https://kns.cnki. net/kcms/detail/34.1294.TN.20230516.1317.002.html), published online May 17 , 2023 Manuscript received: 2023-03-11 Abstract: Emerging Internet of Things (IoT) applications require faster execution time and response time to achieve optimal performance. However, most IoT devices have limited or no computing capability to achieve such stringent application requirements. To this end, compu⁃ tation offloading in edge computing has been used for IoT systems to achieve the desired performance. Nevertheless, randomly offloading ap⁃ plications to any available edge without considering their resource demands , inter-application dependencies and edge resource availability may eventually result in execution delay and performance degradation. We introduce Edge-IoT, a machine learning-enabled orchestration framework in this paper, which utilizes the states of edge resources and application resource requirements to facilitate a resource-aware offloading scheme for minimizing the average latency. We further propose a variant bin-packing optimization model that co-locates applica⁃ tions firmly on edge resources to fully utilize available resources. Extensive experiments show the effectiveness and resource efficiency of the proposed approach. Keywords: edge computing; execution time; IoT; machine learning; resource efficiency Citation (Format 1): AWADA U, ZHANG J K, CHEN S, et al. Machine learning driven latency optimization for Internet of Things applications in edge computing [J] . ZTE Communications, 2023, 21(2): 40 –52. DOI: 10. 12142/ZTECOM.202302007 Citation (Format 2): U. Awada, J. K. Zhang, S. Chen, et al.,“Machine learning driven latency optimization for Internet of Things applications in edge computing,”ZTE Communications, vol. 21, no. 2, pp. 40 –52, Jun 2022. doi: 10. 12142/ZTECOM.202302007. 1 Introduction The Internet of Things (IoT) describes physical devices that are connected to the Internet or networks for the purpose of exchanging and sharing data. IoT enables direct fusion of physical devices into computer sys⁃ tems, resulting in efficiency, more reliable services and eco⁃ nomic benefits without human intervention. However, most IoT devices have limited or no computing capability to meet some application-specific requirements. For example, emerg⁃ ing IoT technologies such as the smart city[1], healthcare-IoT[2], Internet of Vehicles (IoV)[3 –5], connected and autonomous ve⁃ hicles (CAVs) [6], and industry 4.0[7], require substantial re⁃ sources to execute their applications. In addition, most of these applications are structured as a collection of loosely- This work is supported by the National Natural Science Foundation of Chi⁃ na under Grant Nos. 61571401 and 61901416 (part of the China Postdoc⁃ toral Science Foundation under Grant No. 2021TQ0304) and the Innova⁃ tive Talent Colleges and the University of Henan Province under Grant No. 18HASTIT021 . coupled services that communicate with one another and are often latency-sensitive. A conventional approach is to offload these applications to a cloud computing (CC)[8] data center for execution. CC provides an on-demand availability of compute resources over multiple locations, each of which is a data cen⁃ ter. However, a CC data center could be

FAQs

Why is latency optimization critical for IoT applications in edge computing?

IoT applications are often latency-sensitive and require faster execution and response time. Limited device compute makes offloading necessary, and poor scheduling can increase execution delays and degrade performance.

What problem does the Edge-IoT framework address?

Edge-IoT targets inefficient offloading decisions caused by randomly assigning applications without considering resource demands, dependency relationships, and edge resource availability, which can raise average latency.

How does the proposed method reduce average latency and improve resource efficiency?

It uses edge resource states and application requirements for a resource-aware offloading scheme to minimize average latency, and a bin-packing optimization model to co-locate applications effectively on edge resources.

Aug 5, 2026

Upload by Himbo