Motion Constraints and Vanishing Point Aided Land Vehicle Navigation

Area of Science:

• Robotics
• Computer Vision
• Navigation Systems

Background:

• Inertial Navigation Systems (INS) are crucial for ground vehicle navigation but suffer from drift during Global Navigation Satellite System (GNSS) signal outages.
• Aiding sensors are typically used to mitigate INS drift and maintain performance.

Purpose of the Study:

• To enhance the heading and lateral positioning accuracy of INS during GNSS signal outages.
• To propose a novel method utilizing vanishing point (VP) observations from a single forward-looking camera to aid INS.

Main Methods:

• Developed a full motion constraint model for online calibration of the INS frame.
• Integrated a vanishing point (VP) module using parallel lane markings to derive camera-to-road frame attitude.
• Augmented the state-space model with vertical attitude error and incorporated the VP module into an Extended Kalman Filter (EKF) framework.

Main Results:

• Simulations and real-world experiments validated the VP-based method.
• Demonstrated improved heading and cross-track position accuracy compared to methods without VP aiding.
• The proposed method showed robustness and enhanced accuracy when combined with conventional visual odometry.

Conclusions:

• The VP-based method effectively aids INS navigation, particularly during GNSS outages.
• This approach offers a significant improvement in heading and lateral positioning accuracy for ground vehicles.
• The method is compatible with existing visual odometry techniques for enhanced INS performance.