About the project
The aim of this project is to design a human-in-the-loop aerial swarm system to detect and suppress wildfires.
Wildfire burns approximately 420 million hectares of land each year, which has a catastrophic ecological and economic effect around the world. Current wildfire suppression methods are not responsive during the early stages of a fire and are less effective once the fire intensifies. Furthermore, the reliance on manual intervention increases the risk to firefighters, as existing methods lack automation and require direct human involvement.
An alternative approach is to use human-in-the-loop deployment of an aerial swarm that can quickly respond to early identification of incipient-stage fires and suppress the fire before it develops. This project aims to develop a swarm distribution and path planning algorithm that can distributedly find the optimum task allocation and path planning for the swarm to have the maximum effect on the spread of the fire.
You will use multi-agent inverse reinforcement learning to follow the path identified by the human operator, who oversees the overall execution and accepts or rejects plans based on the overall mission strategy. Path planning must take into account environmental limitations, including wind, smoke, and limited communication, which may hinder coordination among the aerial robots.
The project will focus on designing a shared representation of the environment that also accounts for the limited communication payload and the uncertainty of connection links between the agents. There will be an opportunity for you to collaborate with project partners across Europe and the US to develop the detection and suppression strategy.
