Adaptive Cloud-Edge Scheduler Using Lightweight AI Models for Real-Time IoT Streams

Abstract

The rapid growth of Internet of Things (IoT) devices has increased the volume of real-time data generated by sensors, smart devices, industrial controllers, and healthcare monitoring systems. Cloud-centric processing provides scalable computation, but it often introduces high communication latency, bandwidth overhead, and energy cost for delay-sensitive IoT streams. Edge and fog computing reduce response time by moving computa-tion closer to data sources, but efficient workload placement across edge, fog, and cloud resources remains a challenging scheduling problem. This paper proposes an adaptive cloud-edge scheduler using lightweight artificial intelligence models for real-time IoT stream placement. The framework dynamically selects edge, fog, or cloud execution based on stream priority, latency deadline, bandwidth availability, workload complexity, CPU uti-lization, memory utilization, queue length, energy availability, and historical scheduling success. Lightweight models such as Decision Tree, Logistic Regression, Random Forest, and Tiny Neural Network are considered to support fast inference under resource constraints. A latency-energy-aware cost function vali-dates model predictions and improves runtime decision quality. The framework also integrates explainable scheduling decisions and a human-in-the-loop override mechanism for operational governance. Experimental evaluation using a working prototype demonstrates adaptive distribution of IoT streams across edge, fog, and cloud layers, with 11 observed scheduling decisions, 55% success rate, 274.796 ms average latency, and 0.111 J average energy consumption. The results indicate that combining lightweight AI prediction with cost-aware validation improves scheduling flexibility, transparency, and practical applicability in real-time IoT systems.

Introduction

The paper presents an adaptive cloud-edge scheduling framework for real-time Internet of Things (IoT) stream processing. As IoT applications such as smart healthcare, industrial automation, smart cities, and intelligent transportation generate massive volumes of continuous sensor data, processing all data in centralized cloud servers results in high latency, increased bandwidth usage, and greater energy consumption. Although edge and fog computing reduce these limitations by processing data closer to its source, determining the optimal processing location for each data stream remains challenging because network conditions, resource availability, and workload characteristics change dynamically.

To address this challenge, the proposed framework uses a lightweight AI-based scheduler that intelligently decides whether an IoT stream should be processed at the edge, fog, or cloud layer. Unlike static scheduling policies that rely on fixed thresholds, the proposed system adapts its decisions according to real-time resource conditions, latency requirements, energy availability, and workload complexity. It also incorporates a latency-energy-aware cost function, explainable AI, and Human-in-the-Loop (HITL) support to improve transparency, trust, and governance.

The framework consists of six layers: IoT stream generation, edge processing, fog coordination, cloud analytics, AI scheduling, and governance. Incoming IoT streams contain metadata such as priority, workload size, latency deadlines, and resource requirements. The scheduler analyzes multiple features—including stream arrival rate, priority, workload complexity, available bandwidth, network delay, CPU and memory utilization, queue length, energy availability, and historical success rate—to determine the most appropriate processing layer.

The scheduling decision is formulated as a multi-class classification problem, where the AI model selects one of three processing destinations: Edge, Fog, or Cloud. The framework evaluates each decision using mathematical models for total latency, energy consumption, and a multi-objective cost function that combines latency, energy usage, resource utilization penalties, and quality-of-service (QoS) violation penalties. This ensures that scheduling decisions balance performance, efficiency, and resource constraints.

The proposed system also includes stream routing with failover support, allowing workloads to be redirected if the selected processing layer becomes unavailable or overloaded. Additionally, explainable decision traces and HITL override capabilities enable administrators to review, understand, and modify scheduling decisions when necessary.

Conclusion

This paper presented an adaptive cloud-edge scheduler using lightweight AI models for real-time IoT stream processing. The proposed framework dynamically selects edge, fog, or cloud processing based on stream features, resource con-ditions, latency constraints, bandwidth, energy, and model confidence. A latency-energy-aware cost function validates AI predictions, while explainability and HITL support improve transparency and control. The prototype results show that the scheduler can distribute streams across edge, fog, and cloud layers and provide de-cision explanations. The observed 55% success rate, 274.796 ms average latency, and 0.111 J average energy demonstrate the feasibility of the prototype under simulated dynamic con-ditions. The framework is suitable for academic demonstration and can be extended for real IoT deployments. Future work will focus on reinforcement learning-based scheduling, federated learning across edge nodes, Kubernetes-based deployment, digital twin-based scheduling, blockchain-based audit logging, 6G-enabled IoT scheduling, and advanced explainable AI mechanisms.

References

[1] Q. Luo, S. Hu, C. Li, G. Li, and W. Shi, “Resource scheduling in edge computing: A survey,” IEEE Communications Surveys & Tutorials, vol. 23, no. 4, pp. 2131–2165, 2021. [Online]. Available: https://doi.org/10.1109/COMST.2021.3106401 [2] M. Goudarzi, M. Palaniswami, and R. Buyya, “Scheduling iot applications in edge and fog computing environments: A taxonomy and future directions,” ACM Computing Surveys, vol. 55, no. 7, pp. 1–41, 2022. [Online]. Available: https://doi.org/10.1145/3544836 [3] B. Jamil, H. Ijaz, M. Shojafar, K. Munir, and R. Buyya, “Resource allocation and task scheduling in fog computing and internet of everything environments: A taxonomy, review, and future directions,” ACM Computing Surveys, vol. 54, pp. 1–38, 2022. [Online]. Available: https://doi.org/10.1145/3513002 [4] H. Djigal, J. Xu, L. Liu, and Y. Zhang, “Machine and deep learning for resource allocation in multi-access edge computing: A survey,” IEEE Communications Surveys & Tutorials, vol. 24, no. 4, pp. 2449–2494, 2022. [Online]. Available: https://doi.org/10.1109/COMST. 2022.3199544 [5] B. Kar, W. Yahya, Y. Lin, and A. Ali, “Offloading using traditional optimization and machine learning in federated cloud-edge-fog systems: A survey,” IEEE Communications Surveys & Tutorials, vol. 25, no. 2, pp. 1199–1226, 2023. [Online]. Available: https://doi.org/10.1109/COMST.2023.3239579 [6] P. Peng, W. Lin, W. Wu, H. Zhang, S. Peng, Q. Wu, and K. Li, “A survey on computation offloading in edge systems: From the perspective of deep reinforcement learning approaches,” Computer Science Review, vol. 53, p. 100656, 2024. [Online]. Available: https://doi.org/10.1016/j.cosrev.2024.100656 [7] E. Baccour, N. Mhaisen, A. A. Abdellatif, A. Erbad, A. Mohamed, M. Hamdi, and M. Guizani, “Pervasive ai for iot applications: A survey on resource-efficient distributed artificial intelligence,” IEEE Communications Surveys & Tutorials, vol. 24, no. 4, pp. 2366–2418, 2022. [Online]. Available: https://doi.org/10.1109/COMST.2022. 3200740 [8] O. Aouedi, T.-H. Vu, A. Sacco, D. C. Nguyen, K. Piamrat, G. Marchetto, and Q. Pham, “A survey on intelligent internet of things: Applications, security, privacy, and future directions,” IEEE Communications Surveys & Tutorials, vol. 27, no. 2, pp. 1238–1292, 2024. [Online]. Available: https://doi.org/10.1109/COMST.2024.3430368 [9] M. G. S. Murshed, C. Murphy, D. Hou, N. Khan, G. Ananthanarayanan, and F. Hussain, “Machine learning at the network edge: A survey,” ACM Computing Surveys, vol. 54, no. 8, pp. 1–37, 2021. [Online]. Available: https://doi.org/10.1145/3469029 [10] R. Singh and S. S. Gill, “Edge ai: A survey,” Internet of Things and Cyber-Physical Systems, vol. 3, pp. 71–92, 2023. [Online]. Available: https://doi.org/10.1016/j.iotcps.2023.02.004

Copyright

Copyright © 2026 Munesula Venkatesh, Dr. M. Saravanan. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.