Decision Fault Tree Learning and Differential Lyapunov Optimal Control for Path Tracking

Area of Science:

• Robotics and Control Systems
• Artificial Intelligence
• Automotive Engineering

Background:

• Autonomous vehicles face challenges in reliable trajectory tracking amidst varying speeds and road conditions.
• Existing methods often lack robust fault-tolerant trajectory tracking capabilities.
• The need for advanced fault detection and precise path following is critical for safe autonomous operation.

Purpose of the Study:

• To propose a novel method, Differential Lyapunov Stochastic and Decision Fault Tree Learning (DLS-DFTL), for fault detection and trajectory tracking in autonomous vehicles.
• To address limitations in current autonomous vehicle systems regarding fault tolerance and precise path following.
• To enhance the safety and reliability of autonomous driving through improved fault management and trajectory control.

Main Methods:

• Differential Lyapunov Stochastic Optimal Control (SOC) with customizable Z-matrices for precise path tracking and management of noise/faults.
• Development of a recommendation trajectory generation model to support safety justifications, especially for vehicles with low ceilings.
• Implementation of Decision Fault Tree Learning (DFTL) for detecting unexpected deviations caused by faults.

Main Results:

• The DLS-DFTL method demonstrates significant accuracy in fault detection and trajectory tracking.
• Experimental results show a 38% enhancement in fault detection rate compared to state-of-the-art methods.
• The proposed method reduces the loss rate by 14% and achieves 24% faster fault detection times.

Conclusions:

• The DLS-DFTL method offers a robust solution for fault detection and trajectory tracking in autonomous vehicles.
• The approach effectively manages noise and fault issues inherent in localization and path planning.
• The study validates the applicability and superior performance of DLS-DFTL through extensive testing and comparison with existing techniques.