High Precision Outdoor and Indoor Reference State Estimation for Testing Autonomous Vehicles.

Area of Science:

• Automotive Engineering
• Robotics
• Navigation Systems

Background:

• Autonomous driving requires precise vehicle state estimation for testing and validation.
• Global Navigation Satellite Systems (GNSS) offer practical and accurate positioning but are susceptible to signal loss.
• Inertial Navigation Systems (INSs) are crucial but often heavily depend on GNSS availability.

Purpose of the Study:

• To develop a robust methodology for autonomous vehicle navigation that minimizes reliance on GNSS.
• To enhance the accuracy and reliability of vehicle state estimation in challenging environments with intermittent satellite signals.
• To introduce a novel LiDAR-based Positioning Method (LbPM) for indoor navigation and complement existing systems.

Main Methods:

• Implemented machine learning for standstill recognition.
• Developed a mathematical model for horizontation of inertial measurements.
• Utilized statistical filtering for sensor fusion and incorporated outlier and drift detection.
• Introduced a novel LiDAR-based Positioning Method (LbPM) for indoor navigation.

Main Results:

• Demonstrated significant improvement in vehicle state estimation accuracy under adverse conditions, including corrupted or absent correction data and sensor drift.
• Validated the methodology's robustness and accuracy against a state-of-the-art INS with Real-Time Kinematic (RTK) correction.
• Achieved accuracy with the proposed LbPM method comparable to systems utilizing RTK correction.

Conclusions:

• The proposed methodology effectively reduces dependency on GNSS for autonomous vehicle navigation.
• The system provides reliable and accurate vehicle state estimation even with degraded or unavailable satellite signals.
• The novel LbPM offers a viable solution for accurate indoor navigation, complementing outdoor GNSS-based systems.