Improving the Performance of Visual Odometry for Autonomous Navigation in Environments with Low Illumination and Uneven Terrain
Journal
International Journal of Fuzzy Systems
ISSN
1562-2479
2199-3211
Date Issued
2025
Author(s)
Abstract
This study introduces a visual odometry (VO) system designed to overcome challenges in low-light conditions and uneven terrain, particularly for autonomous nighttime navigation. The research aims to improve VO performance where traditional methods struggle with blurred images and unstable tracking in such environments. To achieve this, the system integrates three fuzzy control mechanisms within the ORB-SLAM3 framework, which automatically adjust the feature detection threshold, modulate infrared (IR) headlight brightness, and regulate the robot’s speed using vibration tags linked to an accelerometer that detects uneven surfaces. These vibration data are incorporated into ORB-SLAM3’s Atlas database, allowing the robot to dynamically adjust its velocity during navigation. The system was tested in both indoor and outdoor environments using a differential wheel mobile robot equipped with a calibrated stereo camera and an inertial measurement unit (IMU). Results show significant improvements in positioning accuracy, with a final position error of 0.125 m, a threshold error of 0.304 m, and a matching standard deviation of 30.7 m, while positioning stability improved by 60% during nighttime operations. In conclusion, the proposed system provides an adaptive solution for enabling autonomous robots to stably navigate to target locations in low-light conditions.
Subjects
Autonomous navigation
Controllable infrared illumination array
Vibration tag
Visual odometry
Visual SLAM
SDGs
Publisher
Springer Science and Business Media LLC
Type
journal article