A Testbed for Enabling Intelligence Offloading in Edge Computing for Object Detection

• Reginaldo Luna UFSCar
• Nicole Correa Ramos UFSCar
• Ronan Pereira UFSCar
• Hermes Senger USP

DOI: https://doi.org/10.5753/eradsp.2026.30828

Resumo

The use of Machine Learning models and IoT devices for low-latency and energy-efficient decision-making continues to grow. This paper presents and evaluates a ROS 2-based edge computing framework for distributed You Only Look Once (YOLO) object detection on heterogeneous Single Board Computers for autonomous mobile robotics. Experimental results show that performance depends on hardware roles and scheduling decisions rather than computational power alone. Offloading inference to a GPU server reduced inference latency by approximately 30% and increased throughput by more than one order of magnitude, while the proposed Random Weighted scheduler improved load balancing and increased successful object detections in heterogeneous environments.

Referências

Kashyap, A. and Vidyarthi, A. S. (2022). A load balancing strategy for scheduling of jobs in fog computing environment. In Proceedings of the 2nd International Conference on Advances in Computing, Communication, Embedded and Secure Systems (ICACITE), pages 1152–1157, Greater Noida, India.

Kaur, K., Walia, G. S., and Gill, N. S. (2021). A systematic study of load balancing approaches in the fog computing environment. In Proceedings of the 10th International Conference on System Modeling and Advancement in Research Trends (SMART), pages 111–115, Moradabad, India.

Michaelis, O., Schmid, S., and Mostafaei, H. (2024). L3: Latency-aware load balancing in multi-cluster service mesh. In Proceedings of the 25th International Middleware Conference (Middleware ’24), pages 49–61, New York, NY, USA. Association for Computing Machinery.

Shi, W., Cao, J., Zhang, Q., Li, Y., and Xu, L. (2016). Edge computing: Vision and challenges. IEEE internet of things journal, 3(5):637–646.

Zhang, J., Keramat, F., Yu, X., Hernández, D. M., Queralta, J. P., and Westerlund, T. (2022). Distributed robotic systems in the edge-cloud continuum with ros 2: A review on novel architectures and technology readiness. In 2022 Seventh International Conference on Fog and Mobile Edge Computing (FMEC), pages 1–8. IEEE.

Zhou, Z., Yu, H., Xu, C., Chang, Z., Mumtaz, S., and Rodriguez, J. (2018). Begin: Big data enabled energy-efficient vehicular edge computing. IEEE Communications Magazine, 56(12):82–89.