Back to search

Georgia Institute of Technology

High- and Low-Thrust Early Collision Avoidance Maneuver Guidance for a Markov Decision Process Framework

Abstract

dc:description.abstract

We extend upon a Markov decision process (MDP) framework that autonomously makes guidance decisions for satellite collision avoidance maneuvers (CAMs). In this framework, a reinforcement learning policy gradient (RL-PG) algorithm is used to enable direct optimization of guidance policy using historic CAM data. In addition to maintaining acceptable collision risks, this approach seeks to minimize the average propellant consumption of CAMs by making early maneuver decisions. The CAM is modeled as a continuous state, discrete action and finite horizon MDP, where the critical decision is determining when to initiate the maneuver. The MDP models decision rewards using analytical models of collision risk, propellant consumption, and transit orbit geometry. By deciding to maneuver earlier than conventional methods, the Markov policy effectively favors CAMs that achieve comparable rates of collision risk reduction while consuming less propellant. In this thesis we verify the framework using historical data of tracked conjunction events and conduct an extensive CAM trajectory design study to investigate the influence of high-thrust and low impulsive thrust propulsion, as well as trajectory design on the trained (early maneuver) and cut-off policies. The technical contribution of this thesis work is the implementation in a pre-existing MDP framework for autonomous and early maneuver decision-making of a propellant optimized instantaneous transfer phasing maneuver for high-thrust impulsive satellites, as well as a propellant optimized spiral transfer and coast phasing CAM intended for low-thrust propulsion systems, tested with a low impulsive thrust system for proof of concept. First, we establish a baseline impulsive CAM using a high-thrust phasing maneuver, adapted from the original work. When evaluated on synthetic conjunction events, the trained policy consumes significantly less propellant overall and per maneuver in comparison to a conventional cut-off policy that initiates maneuvers 24 hours before the time of closest approach (TCA). On historical conjunction events, the trained policy consumes more propellant overall but less propellant per maneuver. For both historical and synthetic conjunction events, the trained policy exhibits a conservative nature with high true- and false-positive rates, and lower decision accuracy than the cut-off policy, because as expected, it tends to initiate maneuvers earlier in the conjunction timeline when the state covariance is larger and the risk estimates are less reliable. Additionally, we develop a propellant-optimized impulsive CAM by formulating a constrained optimization over the transit orbit period and number of revolutions, which we solve semi-analytically. We demonstrate that utilizing the minimum-propellant transit period results in a significant reduction in propellant mass consumption across the board without significantly affecting the action distribution, convergence behavior, or decision accuracy of the trained policy. Furthermore, we introduce a preliminary non-instantaneous CAM based on continuous-thrust spiral transfer and coasting dynamics under a two-body model with no perturbations, and we test the model on a low impulsive thrust case study. The trained policy under this low impulsive thrust expends less propellant than the cut-off policy, though at the expense of reduced decision accuracy relative to the impulsive cases, while still maintaining its original conservative nature. Additionally, we develop an optimized spiral and coast CAM by formulating a non-convex propellant minimization problem over the thrust arc and coast arc durations, subject to a phase angle constraint derived from the spiral transfer dynamics. We obtain a semi-analytical approximate solution via a polynomial root solver method and implement it into the MDP framework. We test the CAM on the same low impulsive thrust. The optimized low impulsive thrust policy consumes less propellant than the baseline case, achieves improved decision accuracy (although still below that of the impulsive cases) with continued conservative nature, and exhibits similar action distribution and convergence characteristics to its respective baseline formulation.

Degree

thesis:*
Name thesis:degree_name
Aerospace Engineering, MS
Department thesis:degree_department
Aerospace Engineering
Grantor
Georgia Institute of Technology
Year dc:date.issued
2026

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Schillinger Arana, Lander Wolfgang
Advisor dc:contributor.advisor
  • Li, Sarah
Committee members dc:contributor.committeemember
  • Sankar, Lakshmi
  • Gunter, Brian
  • Lightsey, Glenn
ORCID person:identifier.orcid

Rights

Language dc:language.iso
English

Identifiers

dc:identifier.*
Handle dc:identifier.uri
https://hdl.handle.net/1853/82165

Chain of custody

source
Harvested from
Georgia Tech
Base URL
repository.gatech.edu/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Schillinger Arana, Lander Wolfgang. High- and Low-Thrust Early Collision Avoidance Maneuver Guidance for a Markov Decision Process Framework. Georgia Institute of Technology, 2026. https://hdl.handle.net/1853/82165