Ego-Lane Index Estimation Based on Lane-Level Map and LiDAR Road Boundary Detection.

Area of Science:

• Intelligent Transportation Systems
• Robotics and Autonomous Systems
• Sensor Fusion and Navigation

Background:

• Accurate ego-lane index estimation is critical for intelligent vehicles, particularly in environments with unreliable Global Navigation Satellite System (GNSS) signals.
• Traditional methods struggle with particle degeneration due to high-precision sensor data.

Purpose of the Study:

• To propose an improved particle filter (PF) based ego-lane index estimation approach for urban intelligent vehicles.
• To address particle degeneration issues in particle filter schemes using Light Detection and Ranging (LiDAR) data.

Main Methods:

• Utilized a particle filter (PF) framework initialized and propagated using inertial measurement unit (IMU) and odometry dead reckoning.
• Integrated lane-level map information for particle navigation and Global Navigation Satellite System (GNSS) single-point positioning (SPP) for drift correction.
• Introduced a novel lateral particle shifting step based on LiDAR road boundary detection to mitigate particle degeneration, replacing traditional importance weighting.

Main Results:

• The proposed improved PF scheme accurately estimated the ego-lane index at all time steps during urban expressway testing.
• The traditional PF scheme produced incorrect ego-lane index estimations at certain time steps.
• The novel lateral shifting method effectively handled high-precision LiDAR measurements, preventing particle degeneration.

Conclusions:

• The developed PF scheme with lateral particle shifting provides robust and accurate ego-lane index estimation for intelligent vehicles in urban settings.
• This approach significantly improves upon traditional PF methods, especially in GNSS-challenged environments.
• The method demonstrates the potential for enhanced decision-making and safety in autonomous driving systems.