Balloon Control
LAteral Station-keeping in an Uncertain Flow field
R. Davis Born
Jan to Jun 2018 - accepted to 2019 AIAA SciTech
Abstract
We present a planning algorithm that enables an atmospheric sensing agent, such as a high-altitude weather balloon, to maintain a lateral set-point even though it can only actuate vertically. Using Gaussian Process estimates generated from wind forecast data, a novel method of selecting candidate altitudes for use in planning is presented along with a model-predictive receding horizon tree search algorithm to find the locally optimal sequence of altitudes for station-keeping. We also present a method for opportunistically gathering more samples online to update and improve the wind model without deviating too far from the planned trajectory. These algorithms are shown to outperform naïve and greedy methods in several metrics.
Personal contributions
I personally conducted all the work presented in this paper, including all the associated software development to implement the relevant search/optimization algorithms and simulate the dynamic system for which they were intended. My co-author, Professor Mac Schwager, was my advisor on this project and we worked together to edit the final paper.
Results / deliverables
Conference report & presentation detailing methods & findings
Novel algorithms for station-keeping in a flow field
Software package implementing proposed algorithms
Simulation package for evaluating proposed algorithms