Introduction
Wildfires destroy over 12 million hectares of forest annually, nearly half the size of the UK, causing severe ecological and economic damage.
Current monitoring methods lack the spatial and temporal resolution required for early detection and rapid situational awareness.
This PhD project focuses on developing advanced algorithms for collective perception and prediction in a distributed UAV swarm to enable real-time wildfire monitoring and forecasting. You will explore how a swarm of UAVs can autonomously sense, communicate, and reason about wildfire dynamics to produce actionable information. The project will specifically focus on fusion of distributed observations and interaction with a UAV swarm.
The goal of this project is to enable a swarm to jointly detect, localise, and map fire fronts under uncertainty and partial observability.
Key algorithmic approaches include
Probabilistic State Estimation: Implement distributed filtering methods for fusing noisy observations from multiple UAVs
Collective Perception Algorithms: Design decentralized swarm algorithms that allow the robots to share local detections and collectively estimate fire boundaries in real-time
Multi-Robot Coordination: Apply consensus algorithms and belief propagation to ensure the swarm maintains a coherent, up-to-date situational map despite intermittent communication
Adaptive Sensing Strategies: Use reinforcement learning or adaptive control to allocate the robots' attention to areas of high uncertainty or predicted fire growth
Research Benefits
While the focus of this research will be on developing algorithms using existing data in simulation, the student will have access to multiple small and large aerial platforms to conduct experiments, if interested. There will be the opportunity to participate in networking events and collaborate with a large consortium of world-leading researchers across the UK and US, and if interested, gain hands-on experience with UAVs being developed for wildfire detection and suppression.