The growing complexity of critical infrastructures such as power grids, transportation networks, and water supply has introduced the need for effective monitoring solutions. However, advanced data analytics requires more computational power than resource-constrained edge devices can provide. This paper explores the deployment of AI-driven models in the cloud-edge continuum where sensors connect to edge nodes that forward the data to a powerful central server. The central server handles intensive tasks like training and updating AI models, which are then distributed to edge devices for real-time inference. We handle issues such as resource constraints and the diminishing performance of static models over time. We present approaches to improve response times and data processing by introducing optimized setups and implementing dynamic model updates. Our findings show that integrating AI models with edge devices supported by centralized training can improve real-time monitoring capabilities in resource-constrained environments.

Top