EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A Quad-Objective Edge AI Framework for Real-Time Object Detection and Tracking on Resource-Constrained Embedded Systems

Amulya S

Journal of Intelligent Decision Making and Information Science · 03.10.2026 · 3 (15s) · 2859-2880

DOI: 10.17762/jidmis.v3.4842 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Intelligent Decision Making and Information Science, online erschienen am 03.10.2026. Verfasst von Amulya S. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

One of the biggest challenges for edge deployment of deep learning models on resource-constrained embedded systems is to detect and track objects in real time. This paper introduces an integrated comparative analysis framework based on the four layers: latency, memory footprint, energy dissipation, thermal dissipation are interdependent. We propose Quad-Objective Edge AI Ecosystem (Q-EASE) Framework and validate it with 720 experimental runs on 12 embedded platforms, 3 model families (Convolutional Neural Network (CNN)-based YOLO-NAS, Spiking Neural Network (SNN)-based Spiking YOLO, Transformer-based MobileViT) and 5 public datasets (COCO 2017, VisDrone 2019, MOT17, DanceTrack, UCF101). We present novel contributions such as (a) an algorithmically defined Edge AI Ecosystem architecture, (b) four new optimization algorithms (Dynamic Thermal-Aware Scheduler, Energy-Delay Product (EDP) Optimizer, Adaptive Memory Manager, Latency-Jitter Controller), (c) graphical analysis of trade-off surfaces, Pareto frontiers, and metrics of ecosystem health, as well as (d) design recommendations to achieve 70–95% energy reduction while retaining real-time performance with SNNs. The results demonstrate that the energy-thermal co-optimization is a key differentiator for successful deployments as compared to less successful deployments across the configurations tested (Kendryte K210, YOLO: 8.1/10 health score, 22 hrs. battery; Jetson Nano, YOLO-NAS: 6.1/10 health score, 1.5 hrs. battery, 60% latency increase). On thermally constrained platforms these differences are statistically significant (p

Zitieren

APA

S, A. (2026). A Quad-Objective Edge AI Framework for Real-Time Object Detection and Tracking on Resource-Constrained Embedded Systems. Journal of Intelligent Decision Making and Information Science, 3(15s), 2859-2880. https://doi.org/10.17762/jidmis.v3.4842

Vancouver

S A. A Quad-Objective Edge AI Framework for Real-Time Object Detection and Tracking on Resource-Constrained Embedded Systems. Journal of Intelligent Decision Making and Information Science. 2026;3(15s):2859-2880. doi:10.17762/jidmis.v3.4842

BibTeX
@article{S2026,
  author = {S, Amulya},
  title = {A Quad-Objective Edge AI Framework for Real-Time Object Detection and Tracking on Resource-Constrained Embedded Systems},
  journal = {Journal of Intelligent Decision Making and Information Science},
  year = {2026},
  volume = {3},
  number = {15s},
  pages = {2859--2880},
  doi = {10.17762/jidmis.v3.4842}
}
Kostenlos starten →