Machine learning-aided LiDAR range estimation.

Area of Science:

• Optics and Photonics
• Machine Learning Applications
• Automotive Sensing Technology

Background:

• Automotive light detection and ranging (LiDAR) systems need efficient range estimation.
• Current methods achieve efficiency by limiting the receiver's dynamic range, impacting performance.
• A trade-off exists between efficiency and dynamic range in LiDAR receivers.

Purpose of the Study:

• To propose a novel approach for accurate and computationally efficient range estimation in automotive LiDAR.
• To overcome the dynamic range limitations of current LiDAR receiver technologies.
• To leverage machine learning for improved LiDAR performance.

Main Methods:

• Development of simple yet powerful decision tree ensemble machine learning models.
• Application of these models to automotive LiDAR range estimation.
• Testing and validation of model performance across a significant dynamic range.

Main Results:

• Accurate range measurements were achieved using the proposed machine learning models.
• The models successfully operated across a 45-dB dynamic range.
• The decision tree ensemble approach effectively addressed the efficiency-dynamic range trade-off.

Conclusions:

• Decision tree ensemble models offer a viable solution for enhancing automotive LiDAR systems.
• This machine learning approach enables accurate and efficient range estimation over extended dynamic ranges.
• The findings pave the way for more robust and capable LiDAR technology in autonomous vehicles.