Decision-Making Architectures for Edge AI: Designing Scalable, Low-Latency Systems for Autonomous Intelligence in Distributed and Resource-Constrained Environments
Chen Wei Lun
Department of Computer Science, Southern Taiwan University of Technology, No. 25 Minzu Road, Yongkang District, Tainan, 710012, Taiwan.
Lin Mei Yu
Department of Computer Science, National Yunlin University of Applied Sciences, No. 123 Daxue Road, Douliu City, Yunlin, 640002, Taiwan.
Abstract
Edge AI involves executing AI algorithms on edge devices close to the data source, offering advantages like reduced latency, enhanced privacy, and decreased bandwidth usage. Effective decision-making is crucial in Edge AI for real-time responsiveness, especially in critical applications such as autonomous vehicles and healthcare monitoring. Traditional decision-making models, including rule-based systems and basic machine learning algorithms, often struggle with the dynamic and resource-constrained nature of edge environments. This research aims to explore advanced decision-making techniques leveraging deep learning, reinforcement learning, and federated learning, tailored to the constraints of edge devices. We developed and tested prototypes on actual edge hardware, focusing on computational efficiency, memory usage, latency, and accuracy. Our findings indicate that advanced decision-making architectures can significantly enhance the performance and autonomy of Edge AI systems, paving the way for more efficient, reliable, and intelligent edge applications. This paper provides a comprehensive exploration of these techniques, contributing to the ongoing development and improvement of Edge AI.
