Low-Overhead Scheduling for Real-Time AI Workloads on Multi-Core Edge Chips
Keywords:
Edge Computing, Multi-Core Scheduling, Real-Time Systems, Energy Efficiency, Deep LearningAbstract
When deep neural networks run on heterogeneous edge hardware, they often suffer from contention over shared memory, which frequently violates the strict timing requirements that safety critical systems must guarantee. The mainstream approach in academia to address these scheduling issues currently employs deep reinforcement learning. However, this introduces a new problem: DRL schemes can themselves be relatively heavyweight, and the additional overhead introduced undermines their inherent latency advantages. To alleviate this issue, this study introduces a hybrid framework called Lite Sched to separate computation from real‑time operations. Our system performs offline static analysis and constructs resource occupancy maps for neural network layers, enabling a lightweight, heuristically driven agent to make scheduling decisions with negligible processor overhead. Evaluations on the NVIDIA Jetson Orin platform show that Lite Sched reduces tail latency by 35.7% and improves energy efficiency by 21.4%. Compared with the standard greedy method, our work achieves significant improvements. These results demonstrate that the combination of static analysis and dynamic micro‑scheduling is a reliable way to balance energy and latency trade offs in modern edge computing for resource constrained autonomous industrial systems.
References
Easwaran, A., Yuhas, M., Ramanathan, S., & Samaddar, A. (2024). Real-Time Scheduling for Computing Architectures. In Handbook of Computer Architecture (pp. 127-170). Singapore: Springer Nature Singapore.
Ngo, D., Park, H. C., & Kang, B. (2025). Edge intelligence: A review of deep neural network inference in resource-limited environments. Electronics, 14(12), 2495.
Huang, J., Lin, W., Wu, W., Wang, Y., Zhong, H., Wang, X., & Li, K. (2025). On Efficiency, Fairness and Security in AI Accelerator Resource Sharing: A Survey. ACM Computing Surveys, 57(9), 1-35.
Abdi, A., & Salimi-Badr, A. (2023). ENF-S: An evolutionary-neuro-fuzzy multi-objective task scheduler for heterogeneous multi-core processors. IEEE Transactions on Sustainable Computing, 8(3), 479-491.
Liu, W. (2025). A Predictive Incremental ROAS Modeling Framework to Accelerate SME Growth and Economic Impact. Journal of Economic Theory and Business Management, 2(6), 25–30.
Yu, C., Wang, H., Chen, J., Wang, Z., Deng, B., Hao, Z., ... & Song, Y. (2026). When Rules Fall Short: Agent-Driven Discovery of Emerging Content Issues in Short Video Platforms. arXiv preprint arXiv:2601.11634.
Liu, W. (2025). Few-Shot and Domain Adaptation Modeling for Evaluating Growth Strategies in Long-Tail Small and Medium-sized Enterprises. Journal of Industrial Engineering and Applied Science, 3(6), 30–35.
Lin, A. (2025). Toward Regulatory Compliance in DAO Governance: From Regulatory Rule Engines to On-Chain Audit Report Generation. Journal of World Economy, 4(6), 12-20.
Scrugli, M. A., Meloni, P., Sau, C., & Raffo, L. (2021). Runtime adaptive iomt node on multi-core processor platform. Electronics, 10(21), 2572.
Wang, H., Li, Q., & Liu, Y. (2024). Multi-response Regression for Block-missing Multi-modal Data without Imputation. Statistica Sinica, 34(2), 527.
Wang, H., Sun, W., & Liu, Y. (2022). Prioritizing autism risk genes using personalized graphical models estimated from single-cell rna-seq data. Journal of the American Statistical Association, 117(537), 38-51.
Wu, Y. (2026). A Study on the Impact of Cross-Departmental Data Collaboration on Marketing Campaign Efficiency in Fast-Moving Consumer Goods E-commerce: The Case of PepsiCo (China)’s 7UP and Mirinda Project. Frontiers in Management Science, 5(1), 7-12.
Liu, W. (2025). Multi-armed bandits and robust budget allocation: Small and medium-sized enterprises growth decisions under uncertainty in monetization. European Journal of AI, Computing & Informatics, 1(4), 89–97.
Kohli, P., Jayanth, R., Gupta, N., Fan, H., & Prasanna, V. (2025, September). Performance-Energy Characterization of ML Inference on Heterogeneous Edge AI Platforms. In 2025 IEEE High Performance Extreme Computing Conference (HPEC) (pp. 1-7). IEEE.
Dehghani, A., Fadaei, S., Ravaei, B., & RahimiZadeh, K. (2023). Deadline-aware and energy-efficient dynamic task mapping and scheduling for multicore systems based on wireless network-on-chip. IEEE Transactions on Emerging Topics in Computing, 11(4), 1031-1044.
Yu, C., Li, P., Wu, H., Wen, Y., Deng, B., & Xiong, H. (2024). USM: Unbiased Survey Modeling for Limiting Negative User Experiences in Recommendation Systems. arXiv preprint arXiv:2412.10674.
Li, K., Chen, X., Song, T., Zhou, C., Liu, Z., Zhang, Z., ... & Shan, Q. (2025). Solving situation puzzles with large language model and external reformulation. arXiv preprint arXiv:2503.18394.
Zhu, H., Luo, Y., Liu, Q., Fan, H., Song, T., Yu, C. W., & Du, B. (2019). Multistep flow prediction on car-sharing systems: A multi-graph convolutional neural network with attention mechanism. International Journal of Software Engineering and Knowledge Engineering, 29(11n12), 1727–1740.
Lin, A. (2026). Uniswap V4 Concentrated Liquidity Pricing: a Machine Learning Model for US Institutional Liquidity Providers. Journal of Intelligence and Engineering Technology, 1(1), 19-26.
Wang, J., Kudagama, B. J., Perera, U. S., Li, S., & Zhang, X. (2025). Framework for generating high-resolution Hong Kong local climate projections to support building energy simulations. Physics of Fluids, 37(3).
Liu, Z., Jin, C., Li, S., Li, W., & Wang, J. (2024). Improvement for modeling the damping of the wake oscillator based on the Van der Pol scheme. Physics of Fluids, 36(7).
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Zihe Hao
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
