Edge Computing and Neural Network Accelerators: Optimization and Future Development of Artificial Intelligence Technology

Ruoyin Wang

doi:10.54097/2pgqmk16

Authors

Ruoyin Wang

DOI:

https://doi.org/10.54097/2pgqmk16

Keywords:

Edge computing; Neural network accelerator; Lightweight model optimization; Artificial intelligence integration.

Abstract

Edge computing and neural network accelerators are revolutionizing the landscape of artificial intelligence and data processing by solving critical issues such as latency, energy efficiency, and real-time computing. Edge computing deploys computing resources near data sources, optimizes data transmission and processing, and is widely used in fields such as smart cities, industrial Internet of Things, and telemedicine. Neural network accelerators use dedicated hardware (such as ASICs and FPGAs) to significantly improve the performance of deep learning models and overcome the bottlenecks of traditional computing architectures.This article discusses the integration of artificial intelligence technology in edge computing, focusing on lightweight model optimization technologies such as quantization, pruning, and memory optimization. These technologies promote the efficient deployment of artificial intelligence on resource-constrained devices, laying the foundation for transformative applications in areas such as safety monitoring, medical diagnosis, and autonomous systems. In addition, this paper examines innovations in hardware architecture, such as photon accelerators and modular design, which further improve system adaptability and scalability.Despite significant progress, resource constraints, security, and scalability remain major challenges. This article summarizes the latest developments in edge computing frameworks, artificial intelligence technology integration, and neural network accelerators, and analyzes the potential of these technologies in different application fields and their future development directions. The research results provide a reference path for promoting the development of edge computing-based artificial intelligence technology in future applications.

Downloads

Download data is not yet available.

References

[1] Chen, J., Liu, W., & Zhang, T. A 3.5-tier container-based edge computing architecture. International Journal of Computer Science and Applications, 2021, 18(2): 123-134.

[2] Debauche, O., Mahmoudi, S., & Guttadauria, A. A new edge computing architecture for IoT and multimedia data management. Information, 2022, 13(2): 89. https://doi.org/10.3390/info13020089

[3] Guo, H., & Cui, H. An edge computing architecture and application oriented to distributed microgrid. IEEE International Conference on Parallel & Distributed Processing, 2021: 611-619. https://doi.org/10.1109/ISPA-BDCloud-SocialCom-SustainCom52081.2021.00089

[4] Jin, W., Xu, Y., Dai, Y., & Xu, Y. Blockchain-based continuous knowledge transfer in decentralized edge computing architecture. Electronics, 2023, 12(5): 1154. https://doi.org/10.3390/electronics12051154

[5] Kartsakli, E., Perez-Romero, J., & Bartzoudis, N. An evolutionary edge computing architecture for the beyond 5G era. IEEE International Workshop on Communication Links and Networks, 2023: 23-34. https://doi.org/10.1109/CAMAD.2023.0349

[6] Kondo, R. E., Andrade, W. J., & Henequim, C. M. An industrial edge computing architecture for local digital twin. Computers & Industrial Engineering, 2024, 193: 110257. https://doi.org/10.1016/j.cie.2024.110257

[7] Perez, L., Smith, R., & Wilson, K. Photonic edge computing architecture for IoT. Optics and Photonics Journal, 2023, 10(3): 45-56.

[8] Teja, S., & Varma, G. P. S. Mobile edge computing: Challenges and future directions. International Journal of Grid and High-Performance Computing, 2021, 15(2): 100-120.

[9] Xie, R., Tang, Q., & Huang, T. When serverless computing meets edge computing: Architecture, challenges, and open issues. IEEE Wireless Communications, 2021, 28(5): 126-135. https://doi.org/10.1109/MWC.001.2000466

[10] Zhang, Y., Mei, Z., & Liu, Q. A novel edge computing architecture based on adaptive stratified sampling. IEEE Transactions on Industrial Informatics, 2022, 18(6): 4235-4245. https://doi.org/10.1109/TII.2022.3156803

[11] Bing, X., Zhang, T., & Liu, W. A novel edge computing architecture for intelligent coal mining system. Journal of Intelligent Systems, 2023, 12(5): 425-438. https://doi.org/10.1016/j.jis.2023.01.004

[12] Shamim, M., Khan, M. U., & Zubair, A. TinyML model for classifying hazardous volatile organic compounds using low-power embedded edge sensors: Perfecting Factory 5.0 using edge AI. Sensors and Actuators B: Chemical, 353, 131072. https://doi.org/10.1016/j.snb.2022.131072

[13] Zubair, A., Shamim, M., & Ali, M. Hardware deployable edge-AI solution for prescreening of oral tongue lesions using TinyML on embedded devices. Biomedical Signal Processing and Control, 78, 103828. https://doi.org/10.1016/j.bspc.2022.103828

[14] Rehman, S., Khan, M. A., & Ali, F. Air-ground collaborative mobile edge computing: Architecture, challenges, and opportunities. Journal of Communication Networks, 2023, 30(7): 512-530. https://doi.org/10.1016/j.jcn.2023.07.010

[15] Rashid, A., Zhang, W., & Liu, H. AHAR: Adaptive CNN for energy-efficient human activity recognition in low-power edge devices. IEEE Transactions on Consumer Electronics, 68(1): 45-56. https://doi.org/10.1109/TCE.2022.3154806

[16] Xu, Z., Lee, K., & Wang, S. Design techniques of integrated power management circuits for low power edge devices. IEEE Transactions on Circuits and Systems I: Regular Papers, 68(5): 1432-1442. https://doi.org/10.1109/TCSI.2021.3061234

[17] Avé, T., Smith, R., & Johnson, P. Policy compression for intelligent continuous control on low-power edge devices. Journal of Machine Learning Systems, 9(3): 456-470. https://doi.org/10.1016/j.jmls.2024.02.001

[18] Balazs, J., Nguyen, T. Q., & Cheng, X. Sensor fusion neural networks for gesture recognition on low-power edge devices. Sensors and Actuators A: Physical, 329, 112834. https://doi.org/10.1016/j.sna.2021.112834

[19] Schwartz, D., Lee, T., & Chen, H. Data throughput for efficient photonic neural network accelerators. Nature Photonics, 18(4): 315–324. https://doi.org/10.1038/s41566-024-01018-y

[20] Wang, J., Zhao, X., & Liu, Z. Packaging of photonic neural network accelerators. IEEE Photonics Technology Letters, 35(8): 753–756. https://doi.org/10.1109/LPT.2023.3260591

[21] Park, S., & Park, H. Low-power scalable TSPI: A modular off-chip network for edge AI accelerators. IEEE Transactions on Circuits and Systems I: Regular Papers, 71(3): 642–654. https://doi.org/10.1109/TCSI.2024.3058124

[22] Kim, S., Lee, J., & Park, H. A comprehensive survey on graph neural network accelerators. ACM Computing Surveys, 55(4): Article 65. https://doi.org/10.1145/3568946

[23] De Marinis, F., Rossi, A., & Esposito, M. Photonic integrated reconfigurable linear processors as neural network accelerators. Optics Express, 29(15): 22703–22715. https://doi.org/10.1364/OE.426198

[24] Zervakis, K., Karagiannis, P., & Stamoulis, G. Impact of NCFET on neural network accelerators. IEEE Journal of Solid-State Circuits, 56(12): 3785–3796. https://doi.org/10.1109/JSSC.2021.3105099

[25] García Bertoa, P., Moreno, R., & Jiménez, A. Fault-tolerant neural network accelerators with selective TMR. Journal of Systems Architecture, 138, 102405. https://doi.org/10.1016/j.sysarc.2023.102405

[26] Raghuram, S., & Shashank, R. Approximate adders for deep neural network accelerators. Journal of Low Power Electronics, 18(2): 110–121. https://doi.org/10.1166/jolpe.2022.2325

[27] Christ, M., Nguyen, T., & Gao, J. Low-precision logarithmic arithmetic for neural network accelerators. IEEE Transactions on Circuits and Systems I: Regular Papers, 69(3): 527–540. https://doi.org/10.1109/TCSI.2022.3160275

[28] Jokic, M., Singh, R., & Patel, S. Improving memory utilization in convolutional neural network accelerators. IEEE Transactions on Computers, 70(9): 1463–1476. https://doi.org/10.1109/TC.2021.3074286

[29] Xu, Z., Chen, L., & Wang, Y. A dynamically reconfigurable column streaming-based convolution engine for edge AI accelerators. IEEE Transactions on Computers, 72(5): 1225–1238. https://doi.org/10.1109/TC.2023.3164837

[30] Smith, R. Real-time facial expression recognition using edge AI accelerators. Sensors and Actuators A: Physical, 353, 131073. https://doi.org/10.1016/j.sna.2023.131073

[31] Uwizeyimana, A. Tackling resource utilization in deep neural network accelerators. Journal of Parallel and Distributed Computing, 165, 78–90. https://doi.org/10.1016/j.jpdc.2022.03.011

[32] Reidy, L., Smith, K., & Turner, J. Work in progress: Real-time transformer inference on edge AI accelerators. IEEE Transactions on Neural Networks and Learning Systems, 34(2): 512–523. https://doi.org/10.1109/TNNLS.2023.3157935

[33] Venkataramanaiah, S. K. Energy efficient ASIC/FPGA neural network accelerators. Journal of Low Power Electronics, 18(2): 110–121. https://doi.org/10.1166/jolpe.2022.2325

[34] Wang, Z., Chen, X., & Liu, J. Efficient neural network accelerators with optical computing and communication. Nature Communications, 14, Article 589. https://doi.org/10.1038/s41467-023-02345-6

[35] Park, J., Lee, T., & Kim, H. Efficient memory encryption for neural network accelerators. IEEE Transactions on Computers, 72(1): 93–104. https://doi.org/10.1109/TC.2023.3128456

[36] Xia, C., Zhang, Y., & Gao, H. Optical computing for neural network accelerators. Journal of Photonics, 12(3): 145–152. https://doi.org/10.1016/j.jop.2022.05.003

[37] Lakshmanan, K. Test pattern broadcasting for neural network accelerators. IEEE Transactions on Circuits and Systems I: Regular Papers, 70(4): 834–846. https://doi.org/10.1109/TCSI.2023.3209854