Is this project an undergraduate, graduate, or faculty project?

Graduate

Project Type

individual

Campus

Daytona Beach

Authors' Class Standing

Emre Girgin, Graduate Student

Lead Presenter's Name

Emre Girgin

Lead Presenter's College

DB College of Engineering

Faculty Mentor Name

Cagri Kilic

Abstract

This project proposes a robust framework for enhancing visual-inertial odometry (VIO) in autonomous space surface robots operating in challenging conditions, such as the Moon's dimly lit terrain or Martian dust storms. In such environments, traditional vision-based systems encounter difficulties due to the presence of poor visual features. The proposed system integrates few-shot meta-learning with keypoint detection, enabling real-time adaptation to sparse or obscured visual inputs. This approach leverages minimal training data to identify geometrically consistent landmarks and restore missing keypoints when sensor performance is compromised. The approach is integrated within a tightly coupled inertial-vision framework, which combines probabilistic IMU pre-integration with optimization-based visual odometry. It uses an uncertainty-aware mechanism to dynamically adjust sensor input weights during temporary occlusions (e.g., dust clouds). This innovative integration of meta-learned keypoint recovery and multimodal sensor robustness ensures precise and uninterrupted localization in environments where conventional VIO is ineffective. The efficacy of the framework will be validated through high-fidelity simulations (e.g., lunar and Martian environments). The framework addresses key challenges in extraterrestrial robotic navigation, thereby supporting mission resilience for lunar rovers, Mars explorers, and asteroid prospectors. Furthermore, the methodology is applicable to terrestrial applications with intermittent sensing, such as disaster response and underground exploration, by merging adaptive learning with classical sensor fusion to enable reliable autonomy in unpredictable, sensor-deprived environments.

Did this research project receive funding support (Spark, SURF, Research Abroad, Student Internal Grants, Collaborative, Climbing, or Ignite Grants) from the Office of Undergraduate Research?

No

Share

COinS
 

Meta-Learned Keypoint Recovery for VIO in Space Robotics

This project proposes a robust framework for enhancing visual-inertial odometry (VIO) in autonomous space surface robots operating in challenging conditions, such as the Moon's dimly lit terrain or Martian dust storms. In such environments, traditional vision-based systems encounter difficulties due to the presence of poor visual features. The proposed system integrates few-shot meta-learning with keypoint detection, enabling real-time adaptation to sparse or obscured visual inputs. This approach leverages minimal training data to identify geometrically consistent landmarks and restore missing keypoints when sensor performance is compromised. The approach is integrated within a tightly coupled inertial-vision framework, which combines probabilistic IMU pre-integration with optimization-based visual odometry. It uses an uncertainty-aware mechanism to dynamically adjust sensor input weights during temporary occlusions (e.g., dust clouds). This innovative integration of meta-learned keypoint recovery and multimodal sensor robustness ensures precise and uninterrupted localization in environments where conventional VIO is ineffective. The efficacy of the framework will be validated through high-fidelity simulations (e.g., lunar and Martian environments). The framework addresses key challenges in extraterrestrial robotic navigation, thereby supporting mission resilience for lunar rovers, Mars explorers, and asteroid prospectors. Furthermore, the methodology is applicable to terrestrial applications with intermittent sensing, such as disaster response and underground exploration, by merging adaptive learning with classical sensor fusion to enable reliable autonomy in unpredictable, sensor-deprived environments.

 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.