Development of a Particle Filter-Based Path Tracking Algorithm of Autonomous Trucks with a Single Steering and...

Area of Science:

• Robotics and Automation
• Computer Vision
• Control Systems Engineering

Background:

• Autonomous vehicle and automated guided vehicle (AGV) research focuses on enhancing safety and efficiency.
• Current perception methods using AI and sensors require extensive data, posing limitations.
• Effective path tracking necessitates environmental recognition and robust control strategies.

Purpose of the Study:

• To develop and validate a novel path tracking algorithm for autonomous trucks using a monocular camera.
• To compare the performance of the proposed Linear-Quadratic-Regulator (LQR) based controller against conventional methods.
• To address the limitations of data-intensive AI perception methods in autonomous driving.

Main Methods:

• A particle-filter-based algorithm was employed for target RGB recognition using a monocular camera.
• Path tracking errors were calculated to derive an LQR-based desired steering angle.
• Autonomous truck steering and driving were controlled via pulse-width-modulation (PWM) motors.

Main Results:

• The LQR-based path tracking algorithm was successfully verified in three distinct evaluation scenarios.
• Performance comparison against Sliding Mode Control (SMC) showed the LQR method's superiority.
• The Root Mean Square (RMS) of the lateral preview error for SMC was approximately 18% greater than the LQR-based method.

Conclusions:

• The proposed LQR-based path tracking controller offers improved accuracy and robustness for autonomous trucks.
• This approach provides an effective alternative to data-heavy AI perception systems.
• The study validates the feasibility of monocular camera-based path tracking for autonomous driving applications.