An improved observer design approach for autonomous vehicles using error-based ultra-local model.

Area of Science:

• Control Systems Engineering
• Robotics
• Automotive Engineering

Background:

• Autonomous vehicle systems require accurate state estimation for safe and efficient operation.
• Traditional observer designs often struggle with model uncertainties and nonlinearities inherent in vehicle dynamics.
• Existing methods may not adequately address the complex, real-world conditions encountered by autonomous vehicles.

Purpose of the Study:

• To develop a novel observer design method for autonomous vehicle state estimation.
• To enhance the performance of Linear Parameter Varying (LPV) observers by integrating an error-based ultra-local model.
• To effectively manage unmodeled dynamics and nonlinearities in vehicle models.

Main Methods:

• Combination of the Linear Parameter Varying (LPV) framework with an error-based ultra-local model.
• The error-based ultra-local model is utilized to compensate for uncertainties and nonlinearities.
• Implementation for the specific estimation of lateral velocity in autonomous vehicles.

Main Results:

• Significant improvement in the performance of the LPV-based observer due to the novel design.
• Validation of the observer algorithm's efficiency and operational capability through simulations.
• Demonstration of effectiveness using real-world test measurements from ZalaZone proving ground.

Conclusions:

• The proposed observer design method effectively enhances state estimation for autonomous vehicles.
• The integration of LPV and error-based ultra-local models provides a robust solution for handling model uncertainties.
• The method shows practical applicability, confirmed by both simulation and real-world data.