A Tightly Coupled Visual-Inertial GNSS State Estimator Based on Point-Line Feature

Area of Science:

• Robotics and Autonomous Systems
• Geomatics Engineering
• Computer Vision

Background:

• Visual-Inertial Odometry (VIO) systems provide local pose estimation but suffer from drift.
• Traditional feature-based VIO methods struggle in environments with weak or repetitive textures.
• Global Navigation Satellite System (GNSS) offers global positioning but has limitations in accuracy and real-time performance.

Purpose of the Study:

• To develop a tightly coupled GNSS-VIO system for robust and drift-free state estimation.
• To enhance VIO robustness in challenging environments by incorporating line features.
• To improve the accuracy and real-time capabilities of state estimation by fusing GNSS and VIO data.

Main Methods:

• Proposed a tightly coupled GNSS-VIO system utilizing point-line features for improved environmental structure representation.
• Fused GNSS pseudorange and carrier phase measurements using a carrier phase smoothed pseudorange approach for state estimation.
• Implemented real-time extrinsic parameter calibration between the GNSS receiver and Inertial Measurement Unit (IMU).

Main Results:

• The proposed system demonstrated improved positioning precision compared to traditional methods.
• Experimental results on public datasets confirmed the system's robustness and real-time performance.
• The tightly coupled GNSS-VIO state estimator showed full observability of states in the ECEF frame and consistency.

Conclusions:

• The developed GNSS-VIO system effectively mitigates VIO drift and enhances state estimation robustness.
• The integration of point-line features and smoothed GNSS measurements leads to superior performance in complex environments.
• The system achieves accurate, robust, and real-time state estimation for autonomous navigation applications.