Toward reliable localization: exploring infrared fiducial markers and multi-imu fusion as alternative sensor modalities

Choi, Su-Yeon

Permalink

https://hdl.handle.net/2142/132545 Copy

Description

Title

Toward reliable localization: exploring infrared fiducial markers and multi-imu fusion as alternative sensor modalities

Author(s)

Choi, Su-Yeon

Issue Date

2025-12-05

Director of Research (if dissertation) or Advisor (if thesis)

Bretl, Timothy

Doctoral Committee Chair(s)

Bretl, Timothy

Committee Member(s)

• Golparvar-Fard, Mani
• Tran, Huy
• Stefanescu, Ramona

Department of Study

Aerospace Engineering

Discipline

Aerospace Engineering

Degree Granting Institution

University of Illinois Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

Fiducial Marker, Infrared camera, Localization, Pose Estimation, Multi-IMU, Sensor fusion

Abstract

Reliable localization remains a fundamental challenge for autonomous systems operating in environments where traditional sensors underperform. This dissertation explores two alternative sensor modalities: infrared fiducial markers and multi-IMU integration, as means to enhance localization robustness in GNSS- and vision-degraded settings. Vision-based navigation systems, which commonly rely on visible-spectrum cameras and fiducial markers (e.g., AprilTags or ArUco markers), are widely used for estimating position and orientation. However, their performance degrades significantly under poor lighting conditions, occlusions, motion blur, and adverse weather. These limitations make them unreliable in critical scenarios such as nighttime operations or foggy environments. To overcome these challenges, this work proposes an alternative pose estimation framework based on thermal infrared imagery. In addition, low-cost MEMS IMUs typically used in these systems suffer from high noise and bias instability, limiting the overall accuracy and robustness of the pose estimation. To overcome these challenges, this work proposes two alternative approaches: one based on thermal infrared marker, and another based on fusing measurements from multiple IMUs to improve inertial sensing. In the marker-based approach, a passive infrared fiducial marker is introduced, designed using materials with distinct thermal emissivity properties to remain visible under challenging conditions such as low light, fog, and nighttime. A learning-based detection method is developed by adapting an existing visible-spectrum marker detection algorithm to the infrared domain, addressing challenges such as thermal reflections and partial occlusions. To further enhance detection accuracy, a Pix2Pix image-to-image translation model is used as a pre-processing step to remove thermal reflections from infrared imagery. This pre-processing improves input quality and supports a variety of marker designs. Separately, this work investigates the fusion of measurements from multiple low-cost MEMS IMUs to achieve noise characteristics comparable to those of tactical-grade inertial sensors. The fusion process synchronizes angular velocity and acceleration measurements, and its performance is evaluated through Allan variance analysis and real-world visual-inertial SLAM experiments using a wheeled ground robot. Together, these two approaches offer robust alternatives to conventional sensor modalities, enabling reliable localization in environments where GNSS, visual odometry, or single-IMU setups often fail. Applications include Urban Air Mobility, construction site monitoring, underground operations, and autonomous systems operating in GNSS-denied or visually degraded conditions.

Graduation Semester

2025-12

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/132545

Copyright and License Information

Copyright 2025 Su-Yeon Choi

Owning Collections