WatchEDGE: Smart networking for distributed AI-based environmental control

Abstract

Modern applications that make extensive use of AI and ML require the support of an agile network environment that allows raw data to be processed at the edge, where they are collected, while computing functions both for the training and inference phases need to be efficiently moved from site to site. In this paper we present the WatchEDGE project. WatchEDGE studies an advanced edge-computing architecture supporting AI-based image processing applications distributed on geographically-distant sites, which can be adopted for environmental surveillance in the context of smart agriculture and wildlife protection. WatchEDGE is orchestrated to maximize data processing at the edge: each site (or “island”) is equipped with an edge-computing infrastructure that can be provided by fleets of flying drones (FANET); moreover, in the islands there are smart cameras and smart radars; cameras can be either grounded or mounted on board the UAVs. The islands are interconnected by SD-WAN technology. The project is part of the RESTART program, funded by NextGenerationEU.