Back to search

Georgia Institute of Technology

Autonomous and Robust Monocular Simultaneous Localization and Mapping-Based Navigation for Robotic Operations in Space

Abstract

dc:description.abstract

The theoretical background, the synthesis, and the implementation details of estimation frameworks for target-relative spacecraft rendezvous and proximity operations (RPO) and small body probing and surveying (SBPS) predicated on modern simultaneous localization and mapping (SLAM) are considered. The challenges arising in the application of pure visual monocular SLAM to spacecraft relative navigation by testing an off-the-shelf algorithm, ORB-SLAM, on real satellite servicing image sequences, were identified. It is additionally determined that the inclusion of inertial measurement unit-based (IMU) factors, predominantly used in visual-inertial simultaneous localization and mapping (viSLAM), may not provide observability of the ambiguous scale or of the inertial motion over extended arcs, and moreover would not facilitate the smoothing problem. A comprehensive SLAM framework, predicated on monocular image feature point tracking and sensor fusion for on-the-fly navigation and map building is proposed. The work is contrasted to the state-of-the-art methods which instead exploit stereo imaging. A factor graph approach, allowing for the incorporation of asynchronous measurements of diverse modalities, and the inclusion of kinematic and dynamic constraints, is selected. A new relative dynamics factor predicated on the chaser-target relative orbital mechanics is devised and then augmented with the existing relative kinematics factor of Setterfield et al. to account for non-inertial motion of the target center of mass. AstroSLAM, an algorithm solving for the navigation solution of a spacecraft under motion in the vicinity of a small body by exploiting monocular SLAM, sensor fusion, and RelDyn motion factors, is proposed. The developed motion factor encodes a hybrid inertial rate gyro sensor model and vehicle dynamics model, based on the spacecraft-small-body-Sun system, incorporating realistic perturbing effects, which affect the motion of the spacecraft in a non-negligible manner. The RelDyn factor is readily specialized to the spacecraft rendezvous problem by removing the target gravitational pull variable. The data shows that RelDyn out-performs the state-of-the-art preintegrated IMU accelerometer factors, commonly used in visual-inertial SLAM solutions, in one instance of a legacy NASA small body surveying mission and in one instance of an in-lab-generated dataset. On-the-fly target dynamical parameter estimation, such as the center of mass location, the spin vector, and the gravity parameter, is also demonstrated. An existing robotics procedure, dubbed structure from small-motion (SfSM), is leveraged to tackle the challenge of map initialization with small camera baseline and weak-perspective projection

Degree

thesis:*
Level thesis:degree_level
Doctoral
Department dc:contributor.department
Aerospace Engineering
Grantor dc:publisher
Georgia Institute of Technology
Year dc:date.issued
2024

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Dor, Mehregan
Advisor dc:contributor.advisor
  • Tsiotras, Panagiotis
Committee members dc:contributor.committeemember
  • Lightsey, Glenn
  • Christian, John
  • Frank Dellaert
  • Carlone, Luca

Subjects

dc:subject × 5

Rights

Language dc:language.iso
en_US

Identifiers

dc:identifier.*
Handle dc:identifier.uri
https://hdl.handle.net/1853/75319

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
citation

Dor, Mehregan. Autonomous and Robust Monocular Simultaneous Localization and Mapping-Based Navigation for Robotic Operations in Space. Doctoral thesis, Georgia Institute of Technology, 2024. https://hdl.handle.net/1853/75319