Integration of Edge–AI Into IoT–Cloud Architecture for Landslide Monitoring and Prediction
Amrita Joshi, Saurabh Agarwal, Debi Prasanna Kanungo, Rajib Kumar Panigrahi
Abstract
This article presents the development and first-time implementation of an IoT–edge–AI–cloud architecture in an actual landslide location for real-time monitoring and prediction. The proposed architecture benefits the time-critical landslide application by introducing artificial intelligence (AI) and decision-making at the edge of the network. This architecture can address the issues related to network, data packet drops, and device overload while optimizing energy consumption, response latency, and prediction accuracy, all simultaneously. A data offloading scheme is implemented to address the issue of data-packet drops by the IoT-end nodes. This architecture employs an incremental learning approach that periodically retrains the AI model at the edge using real-time data to optimize the prediction accuracy, thus reducing cloud dependency. Compression techniques are also implemented on the edge server to develop light-weighted AI models that can easily run on resource-constrained edge devices.