Skip to main content
Utah's Foremost Platform for Undergraduate Research Presentation
2021 Abstracts

Decentralized Learning of Search Parameters for Cooperating Vehicles Using Gaussian Process Regression

Presenter: Christine Akagi, College of Engineering, Electrical Engineering
Authors: Christine Akagi
Faculty Advisor: Cammy Peterson, College of Engineering, Electrical Engienering
Institution: Brigham Young University

Unmanned aerial vehicles are capable of working as teams to accomplish a wide variety of mission objectives, such as searching for and tracking targets. This poster presents a method that uses target density knowledge accumulated over the life of the vehicle's mission to adapt the search rewards and drive it to visit target rich areas. The target density information is modelled using a Gaussian process and updated as the vehicles search the space. Unlike typical search algorithms, which reward vehicles to visit areas which haven't been recently viewed, this approach incentivizes visiting areas with a higher probability of targets being present. The target density information is propagated through the vehicle communication network using decentralized consensus filters. Vehicles then choose paths that enhance the understanding of the environment by reducing target density uncertainty and revisiting target rich areas. Through numerical simulations we show that this method provides an accurate estimate of the target densities along a road network.