隨著機器學習模型的演進及其在邊緣運算設備上的部署，機器學習應用變得更加廣 泛。然而，隨著邊緣運算的普及，在這些平台上部署機器學習模型面臨著顯著的挑 戰，尤其是在能源效率方面。本文通過一種動態電壓和頻率調整（DVFS）的創新應 用，解決了邊緣設備中能量管理的關鍵需求。我們進行了一系列測試，發現降低GPU 頻率可以有效節省能耗。與傳統方法持續以峰值頻率運行不同，我們的方法在電力 消耗與處理速度之間實現了最佳平衡，專門針對GPU的並行處理需求。該自適應策略 在Nvidia Jetson Nano上實施，與傳統DVFS設置相比，能耗降低達18%，且僅在最理 想的情況下僅增加了1%的推理時間。此研究延長了邊緣設備的運行壽命，增強了其 在對電力需求緊張的應用中的效用，為在各種領域中廣泛採用可持續且高效的AI技 術鋪平了道路。這些發現的影響深遠，為在資源受限環境中部署AI技術帶來了顯著 的改進。

The evolution of machine learning models and their deployment on edge computing devices expands machine learning applications. As edge computing becomes everywhere, deploying machine learning models on these platforms faces significant challenges, particularly in energy efficiency. This paper addresses the critical need for energy management in edge devices through a novel application of dynamic voltage and frequency scaling (DVFS). By conducting a series of tests, we found that the GPU frequencies can be lowered to save energy consumption. Unlike traditional approaches that continuously operate at peak frequencies, our method optimizes the balance between power consumption and processing speed, targeting the unique needs of parallel processing on GPUs. Implemented on the Nvidia Jetson Nano, our adaptive strategy reduces energy consumption by up to 18% compared to conventional DVFS settings with only 1% of inference time overhead in the best case.. This study extends the operational life of edge devices. It enhances their use in power-sensitive applications, paving the way for the broader adoption of sustainable and efficient AI technologies across various sectors. The implications of these findings are substantial, promising significant improvements in the deployment of AI technologies in resource-constrained environments.

[1] Guo, Tian. "Cloud-based or on-device: An empirical study of mobile deep inference." 2018 IEEE International Conference on Cloud Engineering (IC2E). IEEE, 2018.
[2] Liu, Di, et al. "Bringing AI to edge: From deep learning’s perspective." Neurocomputing 485 (2022): 297-320.
[3] Singh, Raghubir, and Sukhpal Singh Gill. "Edge AI: a survey." Internet of Things and Cyber-Physical Systems 3 (2023): 71-92.
[4] Halawa, Hassan, et al. "Nvidia jetson platform characterization." Euro-Par 2017: Parallel Processing: 23rd International Conference on Parallel and Distributed Computing, Santiago de Compostela, Spain, August 28–September 1, 2017, Proceedings 23. Springer International Publishing, 2017.
[5] Wang, Robert J., Xiang Li, and Charles X. Ling. "Pelee: A real-time object detection [6] Blasco, Jorge, et al. "A survey of wearable biometric recognition systems." ACM Computing Surveys (CSUR) 49.3 (2016): 1-35.
[7] Liu, Shaoshan, et al. "Edge computing for autonomous driving: Opportunities and challenges." Proceedings of the IEEE 107.8 (2019): 1697-1716.
[8] JK Jung. tensorrt_demos. https://github.com/jkjung-avt/tensorrt_demos. 2019.
[9] Mei, Xinxin, et al. "A measurement study of GPU DVFS on energy conservation." Proceedings of the Workshop on Power-Aware Computing and Systems. 2013.
[10] Tang, Zhenheng, et al. "The impact of GPU DVFS on the energy and performance of deep learning: An empirical study." Proceedings of the Tenth ACM International Conference on Future Energy Systems. 2019.