Download pdf
Optimization Strategies for Low-Power AI Models on Embedded Devices
- 1 Washington State University, School of Electrical Engineering & Computer Science, Pullman, WA, United State, 99163
* Author to whom correspondence should be addressed.
Abstract
With the growing demand for IoT devices, developing low-power AI models on embedded systems has become increasingly important. However, the efficient implementation of AI models within the computational and battery limitations of these devices remains a significant challenge. This study addresses how model pruning and quantization compression techniques can reduce power consumption without significantly compromising model accuracy. The research method optimizes the performance of the three-color recognition model, organizes a dataset consisting of red, yellow, and green classification images, and pre-processes them to standardize the resolution and format. The research object is to use pruning and quantization techniques in combination to optimize memory and computational efficiency further. Experimental evaluation was performed on an Arduino Nano 32 with a camera model, TensorFlow Lite for Microcontrollers for deployment, and a power measurement tool to record energy consumption. The results demonstrate that these methods significantly reduce energy consumption while maintaining acceptable accuracy for real-time applications. This study provides practical optimization strategies for deploying TinyML on resource-constrained devices, offering valuable insights for low-power AI development in IoT and edge computing applications.
Keywords
TinyML, ultra-low-power, model optimization, quantization, pruning
[1]. Sabovic, A. Aernouts, M. Subotic, D. Fontaine, J. DePoorter, E. Famaey, J. (2023). Towards energy-aware tinyML on battery-less IoT devices. Internet Things. 22: 100736. https://doi.org/10.1016/j.iot.2023.100736
[2]. Partha, P.R. (2022). A review on TinyML: State-of-the-art and prospects. Journal of King Saud University-Computer and Information Sciences. 34: 1595-1623. https://doi.org/10.1016/j.jksuci.2021.11.019
[3]. n.d. (2024). TinyML: Exploring the world of tiny machine learning. LinkedIn. https://www.linkedin.com/pulse/tinyml-exploring-world-tiny-machine-learning-scdte/#:~:text=The%20essence
[4]. Youdi, G. Guangzhen, L. Yunzhi, X. Rui, L. Lingzhong, M. (2023). A survey on dataset quality in machine learning. Information and Software Technology. 162: 107268. https://doi.org/10.1016/j.infsof.2023.107268
[5]. Braka, O. (2023). On Common Split for Training, Validation, and Test Sets in Machine Learning. https://pub.towardsai.net/breaking-the-mold-challenging-the-common-split-for-training-validation-and-test-sets-in-machine-271fd405493d
[6]. n.d. (2024). Understanding the difference: Structured vs. unstructured neural pruning. Medium. https://medium.com/@agi_63938/understanding-the-difference-structured-vs-unstructured-neural-pruning-76292b9384b7#:~:text=Unstructured%20pruning%20removes%20individual%20weights%20from%20a%20neural, zeros.%20Unstructured%20pruning%20is%20easier%20than%20structured%20pruning.
[7]. Francisco, C. (2021).TinyML models: What’s happening behind the scenes. https://medium.com/marionete/tinyml-models-whats-happening-behind-the-scenes-5e61d1555be9
[8]. Bofeng, J. Jun, C. Yong, L. (2023). Single-shot pruning and quantization for hardware-friendly neural network acceleration. Engineering Applications of Artificial Intelligence. 126: 106816. https://doi.org/10.1016/j.engappai.2023.106816
[9]. n.d. (2024). Comparison of Fixed-point Versus Floating-point Quantization in Neural Networks. https://peerdh.com/blogs/programming-insights/comparison-of-fixed-point-versus-floating-point-quantization-in-neural-networks
[10]. Kartik, T. (2024). A Guide to Quantization in LLMs. Symbl.ai. https://symbl.ai/developers/blog/a-guide-to-quantization-in-llms/.
[11]. Ummara, B. Mahrukh, M. Dilshad, S. Muhammad, F.U.B. Ali, H. Mustansar, A.G. Arshad, A.K. Wadood, A. (2024). Advances in Pruning and Quantization for Natural Language Processing. IEEE Access. 12: 139113 - 139128. https://ieeexplore.ieee.org/document/10685352
Cite this article
Zhen,T. (2025). Optimization Strategies for Low-Power AI Models on Embedded Devices. Applied and Computational Engineering,133,38-45.
Data availability
The datasets used and/or analyzed during the current study will be available from the authors upon reasonable request.
Disclaimer/Publisher's Note
The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of EWA Publishing and/or the editor(s). EWA Publishing and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
About volume
Volume title: Proceedings of the 5th International Conference on Signal Processing and Machine Learning
© 2024 by the author(s). Licensee EWA Publishing, Oxford, UK. This article is an open access article distributed under the terms and
conditions of the Creative Commons Attribution (CC BY) license. Authors who
publish this series agree to the following terms:
1. Authors retain copyright and grant the series right of first publication with the work simultaneously licensed under a Creative Commons
Attribution License that allows others to share the work with an acknowledgment of the work's authorship and initial publication in this
series.
2. Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the series's published
version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgment of its initial
publication in this series.
3. Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and
during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See
Open access policy for details).