Road Surface Classification Using a Deep Ensemble Network with Sensor Feature Selection.

Area of Science:

• Computer Science
• Artificial Intelligence
• Robotics

Background:

• Deep learning is crucial for autonomous systems.
• Identifying road conditions is vital for vehicle navigation.
• Sensor data integration presents challenges in autonomous driving.

Purpose of the Study:

• To propose a deep learning network for road identification using diverse sensor data.
• To enhance autonomous vehicle perception capabilities.
• To optimize sensor data utilization through feature selection.

Main Methods:

• Utilized Long-Short Term Memory (LSTM) units and ensemble learning for network architecture.
• Implemented a feature selection technique to exclude redundant sensor data.
• Conducted real-world vehicle experiments for training and validation.

Main Results:

• The proposed deep learning network achieved a 94.6% classification accuracy on test data.
• Performance was validated across four distinct test road environments.
• Feature selection effectively reduced data complexity without compromising accuracy.

Conclusions:

• The developed deep learning model demonstrates high accuracy in road identification for autonomous vehicles.
• The integration of LSTM and ensemble learning offers a robust approach for sensor data fusion.
• This method contributes to safer and more reliable autonomous driving systems.