Control of Systems in Unknown Environments: Autonomous Control for High-Altitude Balloon
Loading...
Authors
Harry, Telema
Date
2025-11-03
Type
thesis
Language
eng
Keyword
Data-Driven Control , Output Regulation of Nonlinear System , High Altitude Balloon , Station-Keeping , Extremum-Seeking Control
Alternative Title
Abstract
The control of systems operating in unknown environments, characterized by model uncertainties, unknown disturbances, and reference signals to be tracked, is a challenging problem in many engineering applications. Output regulation theory is the standard framework for addressing such problems; however, most existing solutions assume that the control direction is known \emph{a priori}. In many practical engineering applications, particularly those with significant model uncertainties and limited state measurements, this assumption does not hold. These limitations motivate the development of adaptive and data-driven control strategies capable of operating reliably under unknown conditions.
This thesis addresses these challenges through two primary contributions. First, we develop a novel controller that integrates a model-free extremum-seeking control (ESC) scheme with an internal model generator. Through rigorous analysis and simulation studies, we demonstrate that the proposed controller solves the output regulation problem for a class of second-order nonlinear systems with unknown control direction.
Second, since the analytical solution of the regulator equations is often difficult to obtain even for simple dynamics, we propose a data-driven learning framework that combines classical control theory with supervised machine learning. This hybrid approach enables the approximation of internal model dynamics at steady state, providing an effective means of implementing the internal model principle without explicit knowledge of the plant or exosystem models.
Finally, we investigate the navigation and station-keeping of unmanned High-Altitude Balloons (HABs) operating in the Earth’s stratosphere under unknown and time-varying wind dynamics. An intermittent dual-mode ESC algorithm is developed to achieve real-time path planning and station-keeping using only locally available data. Simulation results based on real atmospheric wind fields from NOAA show that the proposed approach can steer a balloon across targeted regions and maintain station-keeping performance without reliance on historical wind predictions or explicit aerodynamic modelling.