Swarm-based wildfire monitoring and mitigation

: Using swarm intelligence to solve the global firefighting problem

Abstract

Swarms of high payload uncrewed aerial vehicles (UAVs) can be used to assist firefighters in wildfire operations. Wildfires are increasing due to the impact of climate change. Thus, new tools need to be developed to assist firefighters. Swarm intelligence (SI) can shape adaptable and resilient systems that require little to no intervention by users to achieve a task. At the same time, UAV hardware has evolved, achieving large payload transportation capabilities. Although many developers understand this potential, there is a gap between what users require and the proposed systems.
To remove this gap, we started our work by understanding the complex world of wildfire operations. Firefighters from all over the world were contacted to take part in mutual shaping studies, where developers and end users cooperate to design new systems. Through this study, users explained how they imagined the use of a swarm of high-payload UAVs in firefighting operations. As a result of this process, we focus on a scenario where a pre-deployed swarm is used to detect wildfires early before they develop into large firefronts and mitigate them in cooperation with firefighters in the field.
As field deployments can be very versatile, the swarm will need to adjust its behaviour based on variations such as the change in the number of agents in the swarm or in the number of fires that appear. To achieve this, decentralised bio-inspired swarm intelligence algorithms were created and tested. A novel algorithm named dynamic space partition was developed to dynamically partition an area based on the number of agents so that the swarm can identify wildfire locations. Since the system is designed to mitigate early-stage wildfires, the control strategies were adjusted to engage different types of fires. Different search patterns were created and tested as well as cooperative responses by the swarm.
Swarms need to be tested and deployed in the field. To achieve this, users and developers need to interact with the deployed systems. We therefore developed a user interface (UI) to test, monitor and deploy aerial swarms. The system can create digital twins of aircraft that can also interact with physical aircraft allowing the creation of a mixed-reality swarm. The system was tested in field trials where three digital twins and three physical aircraft were controlled. Tests demonstrate that this system can deploy swarms in the field and enable interaction with the aircraft to modify their behaviour.

Date of Award	13 May 2025
Original language	English
Awarding Institution	University of Bristol
Supervisor	Tom S Richardson (Supervisor) & Sabine Hauert (Supervisor)