Optimized LiDAR target for benchmarking depth measurement accuracy.

Area of Science:

• Robotics and Autonomous Systems
• Sensor Technology
• Metrology

Background:

• Autonomous vehicles rely heavily on Light Detection and Ranging (LiDAR) for 3D perception and depth measurement.
• Verifying LiDAR depth accuracy is crucial for safe autonomous driving, yet benchmark target optimization remains underexplored.
• Existing methods lack sufficient explainability and optimization for robust LiDAR performance evaluation.

Purpose of the Study:

• To propose and validate an optimized LiDAR target for enhanced depth measurement accuracy verification.
• To improve the distinguishability of return signal quality for LiDARs under test.
• To provide a quantifiable method for assessing LiDAR performance differences.

Main Methods:

• Utilized a LiDAR simulator to model scanning errors in relation to signal quality.
• Developed a "disparity function" to quantify differences in depth measurement errors.
• Designed and fabricated an optimized target shape using 3D printing based on simulation results.

Main Results:

• The optimized target significantly increased the difference in depth measurement errors between two LiDAR sensors (over sevenfold compared to a planar target).
• The proposed target effectively prevented confounding error sources from masking performance differences.
• Demonstrated improved reliability in distinguishing LiDAR performance through signal quality metrics.

Conclusions:

• The optimized LiDAR target provides a more effective and reliable method for verifying depth measurement accuracy in autonomous vehicle sensors.
• The "disparity function" offers a valuable tool for quantifying LiDAR performance and guiding target optimization.
• This research contributes to the development of more robust benchmarking for autonomous driving systems.