Author ORCID Identifier

https://orcid.org/0000-0002-6483-8385

Semester

Summer

Date of Graduation

2026

Document Type

Dissertation

Degree Type

PhD

College

Statler College of Engineering and Mineral Resources

Department

Mechanical and Aerospace Engineering

Committee Chair

Hang Woon Lee

Committee Co-Chair

Piyush Mehta

Committee Member

Zeyu Liu

Committee Member

Jason Gross

Committee Member

Guilherme Augusto Silva Pereira

Committee Member

Yu Gu

Abstract

Earth observation satellites (EOSs) are crucial in space operations, most especially in monitoring various planetary phenomena that exist on various spatial and temporal scales. In general, EOSs are the most flexible method used to monitor such phenomena, allowing frequent observations and the use of a variety of observation instruments. An improvement over monolithic satellite systems is the utilization of a constellation of cooperative satellite systems, further reducing revisit times and increasing coverage of target objectives. Constellations of EOSs are often assumed to employ nadir-directional observation techniques, limiting the visible time windows (VTWs) of targets, the time during which said targets are available to the constellation for observation, thus limiting the effectiveness of each EOS. This limitation can be alleviated by incorporating concepts of operation (CONOPS) that extend the VTWs during which targets can be observed from orbit. One such CONOPS is satellite agility, the ability of EOSs to perform slewing operations, thereby changing the angle at which a target can be viewed and potentially extending or opening new VTWs. Satellite agility has been more commonly used in recent years within EOS constellations to increase overall operational capacities. Separately, the emerging CONOPS of constellation reconfigurability, leveraging the ability of satellites to perform orbital maneuvers as a component of nominal operations, has been established as the next paradigm of EOS operations. Constellation reconfigurability enables more optimal constellation configurations to be achieved at multiple opportunities within a mission horizon, further improving the quantity and quality of VTWs. This dissertation develops key methodologies and operational frameworks regarding constellation reconfigurability to improve EOS systems as a whole within multiple contexts, progressing the technology closer to implementation reality.

The enclosed work has achieved multiple objectives related to the study of various implementations of constellation reconfigurability for targeted dynamic planetary phenomena. First, a thorough comparative analysis between the performance of the CONOPS of satellite agility and constellation reconfigurability relative to baseline nadir-directional observations was conducted with respect to the VTWs of 100 historical tropical cyclones (TCs). The results of the comparative analysis indicate that implementing constellation reconfigurability with plane-change maneuvers outperforms satellite agility in the majority of TCs analyzed. Second, a novel mixed-integer linear programming formulation of the Reconfigurable Earth Observation Satellite Scheduling Problem (REOSSP), incorporating orbital maneuver capabilities into the nadir-directional EOSSP, was developed to optimally solve the deterministic scheduling problem for given parameters. Simultaneously, a rolling horizon procedure (RHP) solution method was developed to mitigate computational intractability. The performance of the REOSSP is benchmarked against the EOSSP, which serves as a baseline nadir-directional solution, using a set of random instances in which problem characteristics are varied and a case study of real-world data, including a historical TC path. The results of these experiments demonstrate the value of constellation reconfigurability in the context of the EOSSP, yielding solutions with improved performance, while RHP reduces computational runtime for large-scale REOSSP instances. Third, the REOSSP was implemented within an autonomous algorithmic framework for the detection and subsequent scheduling of operations related to active wildfires. Within this framework, various modules are employed to enable operations, including a Detection Module that leverages convolutional neural networks to detect active wildfires in simulated satellite imagery, a Multi-Pass Confidence Module that statistically filters out false-positive wildfire detections, and a Schedule Module that uses the REOSSP and EOSSP interchangeably. The modular framework is applied in experiments with historical wildfire events using real-world operational parameters, demonstrating its realism and effectiveness in automating the detection and scheduling pipeline in EOS systems. Fourth, since all previous objectives concern deterministic settings, an exploration of stochastic target characteristics was conducted to account for uncertainties in EO tasks and planetary phenomena objectives. A stochastic variant of the Multistage Constellation Reconfiguration Problem (MCRP) was formulated and solved primarily using Stochastic Dual Dynamic Integer Programming. Additional common stochastic problem-solving techniques are presented for a comparison of solution quality in the presence of uncertainty. Two computational experiments are conducted concerning stochastic target properties to demonstrate the effectiveness of each solution approach; the first employs random orbital targets, while the second concentrates on simulated hurricanes. The stochastic MCRP accounts for target stochasticity while obeying visible time windows and maneuver feasibility.

Overall, this dissertation employs the concepts of mixed integer linear programming, model predictive control, convolutional neural networks, Bayesian statistics theory, scheduling techniques, stochastic problem-solving techniques, and stochastic dual dynamic integer programming to achieve implementation. Additionally, experiments are conducted using industry-level programming software and commercial optimization software to obtain provable solutions and overarching results. This exploration of constellation reconfigurability in application to EOSs encompasses VTW performance, constellation scheduling, real-world sequence operations, and optimization of operations under stochastic uncertainty, demonstrating that constellation reconfigurability yields higher performance than nadir-directional and/or agile operations across all cases.

Download

Available for download on Saturday, July 31, 2027

Share

COinS