Past, Present, Future: A Comprehensive Exploration of AI Use Cases in the UMBRELLA IoT Testbed

TL;DR

This paper presents UMBRELLA, a large, open-access IoT testbed designed to enable real-world AI-enabled IoT research through a System-of-Systems approach. It details four AI use cases implemented on UMBRELLA—streetlight monitoring, a data-driven digital twin for building environments, a large-scale federated learning framework, and an intrusion detection system for IoT edge containers—highlighting concrete architectures, data flows, and performance outcomes. The work demonstrates how containerized AI applications, edge and cloud resources, and a unified backend support rapid development, validation, and deployment of AI in distributed IoT settings. It also discusses future smart city and multi-robot crowdsensing scenarios, emphasizing semantic communications and multi-agent planning, and argues for a dedicated MLOps platform to automate AI pipelines and establish trust across the ecosystem.

Abstract

UMBRELLA is a large-scale, open-access Internet of Things (IoT) ecosystem incorporating over 200 multi-sensor multi-wireless nodes, 20 collaborative robots, and edge-intelligence-enabled devices. This paper provides a guide to the implemented and prospective artificial intelligence (AI) capabilities of UMBRELLA in real-world IoT systems. Four existing UMBRELLA applications are presented in detail: 1) An automated streetlight monitoring for detecting issues and triggering maintenance alerts; 2) A Digital twin of building environments providing enhanced air quality sensing with reduced cost; 3) A large-scale Federated Learning framework for reducing communication overhead; and 4) An intrusion detection for containerised applications identifying malicious activities. Additionally, the potential of UMBRELLA is outlined for future smart city and multi-robot crowdsensing applications enhanced by semantic communications and multi-agent planning. Finally, to realise the above use-cases we discuss the need for a tailored MLOps platform to automate UMBRELLA model pipelines and establish trust.

Figures (4)

• Figure 1: UMBRELLA SoS architecture overview with support of multiple sub-system testbeds.
• Figure 2: (a) shows an UMBRELLA node with the enclosure open, showing modules for 1) mothership, 2) edge, 3) ambient sensing, 4) vacant ambient sensing expansion pod, and 5) the RPi Camera; (b) indicates the connection method of 1) mothership pod, 2) edge computing pod and 3) ambient sensing pod within a node.
• Figure 3: This figure illustrates an overview of the UMBRELLA AI use cases presented in this paper, where (a) refers to the street lighting remote monitoring; (b) is digital twin for building environment sensing; (c) depicts the FL trail on UMBRELLA node; (d) shows the intrusion detection on UMBRELLA edge devices.
• Figure 4: Preliminary results using four connected UMBRELLA IoT devices and containerized FL system.