Robust Edge AI for Real-Time Industry 4.0 Applications in 5G Environment
Xiaofeng Zou, Kenli Li, Joey Tianyi Zhou, Wei Wei, Cen Chen
Abstract
Internet of Things (IoT) technologies, including drones, can efficiently capture industrial data, promoting the fourth industrial revolution, Industry 4.0. Moreover, as the 5G technologies evolve, Edge AI can push the AI programs from the remote cloud to the network edges close to end devices, enabling reliable and low-latency intelligent services. Compared with traditional applications, Industry 4.0 applications require more accuracy and lower latency. Most importantly, the robustness of Edge AI system is also critical for Industry 4.0 applications. In this work, we propose a robust Edge AI system for real-time industry 4.0 applications. Our proposed robust AI system can conduct model combination design and model deployment design based on the demands of applications, for example, application accuracy and application latency. Our system is also robust to physical system failures and resumes running intermediately when physical system failures occur.