Predicting driving comfort in autonomous vehicles using road information and multi-head attention models

Area of Science:

• Automotive Engineering
• Artificial Intelligence
• Robotics

Background:

• Driving comfort is vital for autonomous vehicles.
• Current methods focus on local path planning, neglecting macroscopic factors like traffic and road conditions.
• Complex traffic scenarios can lead to emergency braking, negatively impacting comfort.

Purpose of the Study:

• To develop a methodology for predicting driving comfort using road information.
• To integrate driving comfort prediction into global path planning for autonomous vehicles.
• To investigate the interplay between macroscopic factors and driving comfort.

Main Methods:

• Established a road information-driving comfort dataset.
• Devised prediction models utilizing a multi-head attention mechanism.
• Applied the model to global path planning and conducted tests.

Main Results:

• The developed model accurately predicts driving comfort based on road information.
• Vehicles following the optimized path demonstrated a reduction in jerk.
• The integration of comfort prediction with global path planning was validated.

Conclusions:

• Leveraging road information for driving comfort prediction is effective.
• Integrating comfort prediction into global path planning improves autonomous navigation.
• This research offers a valuable framework for enhancing autonomous driving systems.