Data-Driven Steering Dynamics Modeling and Steering Angle Tracking Control for Self-Driving Vehicles: Simulation and...

Area of Science:

• Robotics and Control Systems
• Automotive Engineering
• Machine Learning for Autonomous Systems

Background:

• Autonomous vehicles (AVs) require precise steering control for safety and performance.
• Accurate steering dynamics models are essential for reliable AV operation.
• Industry and academia show growing interest in AVs for enhanced transportation.

Purpose of the Study:

• To develop a data-driven modeling framework for AV steering dynamics.
• To create and evaluate data-driven and PID steering angle tracking controllers.
• To validate the proposed models and controllers using simulated and real-world data.

Main Methods:

• Utilized simulated and real-world driving data from a 2025 Nissan Leaf SV Plus with a drive-by-wire system.
• Developed a data-driven steering dynamics model.
• Implemented and tested a data-driven steering angle tracking controller and a proportional-integral-derivative (PID) controller.

Main Results:

• The PID controller achieved a steady-state tracking root-mean-square error of 1.23° over a ±450° range.
• The neural controller exhibited characterized direction asymmetry in closed-loop tracking.
• Demonstrated the deployability of the data-driven model and controllers on the test vehicle.

Conclusions:

• The proposed data-driven steering dynamics model and control framework are accurate and simple.
• These methods can be applied to the development of next-generation autonomous driving systems.
• Further refinement of the neural controller is suggested to address identified direction asymmetry.