A Lyapunov Optimization-Based Approach to Autonomous Vehicle Local Path Planning.

Area of Science:

• Robotics
• Artificial Intelligence
• Control Systems

Background:

• Autonomous vehicles (AVs) promise enhanced safety and efficiency.
• Achieving full autonomy (SAE Level 5) hinges on effective path planning.
• Current path planning methods face challenges in computational complexity and safety.

Purpose of the Study:

• To introduce a novel Lyapunov Optimization (LO) approach for local path planning in AVs.
• To evaluate the performance of the LO method against conventional techniques.
• To assess the feasibility of a vision-only system for AV path planning.

Main Methods:

• A novel Lyapunov Optimization (LO) model was developed for AV local path planning.
• The LO model was benchmarked against Model Predictive Control and a sampling-based approach.
• An AV prototype utilized a vision-only system for object detection and data collection in Norman, Oklahoma.

Main Results:

• The proposed LO strategy achieved at least a 20-fold reduction in computation time compared to baseline methods.
• Performance was evaluated based on path smoothness, safety, and computation time.
• The vision-only approach proved effective for real-world applicability and cost reduction.

Conclusions:

• Lyapunov Optimization presents a highly efficient solution for AV local path planning.
• The vision-only implementation demonstrates practical viability for autonomous driving systems.
• This research contributes to advancing the safety and computational efficiency of autonomous vehicles.