Cooperative and Distributed Algorithms for Dynamic Fire Coverage using a Team of UAVs

Abstract

Recent large wildfires in the United States and subsequent damage that they have caused have increased the importance of wildfire monitoring and tracking. However, human monitoring on the ground or in the air may be too dangerous and therefore, there needs to be alternatives to monitoring wildfires. Unmanned Aerial Vehicles (UAVs) are currently being considered and used for applications such as reconnaissance, surveying, and monitoring in the spatial-time domain because they can be deployed in teams remotely to gather information and minimize the harm and risk to human operators. UAVs have been previously used in this problem domain to track and monitor wildfires with approaches such as potential fields and reinforcement learning. In this thesis, we aim to look at a team of UAVs, in a distributed approach, over an area to maximize the sensor coverage in dynamic wildfire environments and minimize the energy consumption of deployed UAVs in a network. The work implements and compares an implementation of Deb's NSGA-II to optimize potential fields, Experience Replay with Q-learning, a Deep Q-Network (DQN), and a Deep Q-Network with a state estimator (autoencoder) to track and cover wildfires. The application of this work is a not a final suggestion or an absolute solution for wildfire monitoring and tracking but instead compares the methods to declare the most promising method for future work and research.