GPS/VIO integrated navigation system based on factor graph and fuzzy logic.

Area of Science:

• Robotics and Autonomous Systems
• Navigation and Positioning Technologies
• Computer Vision and Sensor Fusion

Background:

• Global Positioning System (GPS) and Inertial Navigation System (INS) integration faces accuracy and reliability challenges, especially during GPS signal outages.
• Accumulated errors in traditional GPS/INS systems limit long-term navigation solutions.
• Visual-Inertial Odometry (VIO) offers complementary data but requires robust feature handling.

Purpose of the Study:

• To develop a novel graph-based method integrating GPS and VIO for improved navigation and positioning accuracy.
• To introduce Adaptive Feature-Flow Fusion for robust motion estimation and feature extraction in diverse environments.
• To mitigate the accuracy and reliability concerns associated with GPS-dependent navigation systems.

Main Methods:

• A graph-based approach with a dynamically adjustable fuzzy window was employed for GPS/VIO integration.
• Adaptive Feature-Flow Fusion was developed for enhanced motion estimation and feature extraction.
• Data preprocessing, including image optimization and IMU noise reduction, was incorporated.

Main Results:

• The proposed GPS/VIO system demonstrated significant accuracy improvements, outperforming the Extended Kalman Filter (EKF).
• Accuracy enhancements of 84.59% and 88.806% were achieved during GPS outages of 10s and 27s, respectively.
• Predefined trajectory accuracy was improved by 44.7% through data preprocessing techniques.

Conclusions:

• The novel graph-based GPS/VIO system effectively enhances navigation and positioning accuracy, particularly during GPS outages.
• Adaptive Feature-Flow Fusion provides robust performance in varying feature environments.
• The system's real-time performance and accuracy improvements validate its potential for autonomous applications.