Self-Driving Car Location Estimation Based on a Particle-Aided Unscented Kalman Filter

Area of Science:

• Robotics and Autonomous Systems
• Sensor Fusion
• Navigation and Localization

Background:

• Self-driving car operation relies heavily on accurate localization.
• Noisy global positioning system (GPS) signals and urban multipath effects degrade localization performance.
• A practical and precise localization approach is essential for autonomous vehicle safety and reliability.

Purpose of the Study:

• To develop a novel sensor fusion approach for enhanced self-driving car localization.
• To address the limitations of traditional localization methods in challenging urban environments.
• To improve the precision and stability of vehicle positioning.

Main Methods:

• Proposed a particle-aided unscented Kalman filter (PAUKF) algorithm for vehicle localization.
• Utilized the unscented Kalman filter to update vehicle state, accounting for motion models and non-Gaussian noise.
• Integrated a particle filter with onboard sensors and high-definition (HD) maps for refined position measurement.

Main Results:

• The PAUKF algorithm demonstrated superior precision in localization compared to previous methods.
• The proposed method achieved comparable stability in localization performance.
• Simulations validated the effectiveness of the sensor fusion approach in diverse urban scenarios.

Conclusions:

• The developed PAUKF algorithm offers a robust solution for self-driving car localization.
• Sensor fusion effectively mitigates the impact of noisy GPS signals and multipath interference.
• The approach provides a significant advancement in autonomous vehicle navigation accuracy and reliability.