Research on Trajectory Planning of Autonomous Vehicles in Constrained Spaces.

Area of Science:

• Robotics
• Artificial Intelligence
• Computer Science

Background:

• Autonomous vehicles require efficient trajectory planning for safe navigation.
• Complex and constrained environments pose significant challenges to existing planning algorithms.
• Current methods may lack the adaptability and smoothness needed for real-world deployment.

Purpose of the Study:

• To enhance the hybrid A-star algorithm for improved trajectory planning in complex environments.
• To develop a method that increases planning efficiency and ensures smooth vehicle motion.
• To address the limitations of current trajectory planning techniques in challenging scenarios.

Main Methods:

• An adaptive node expansion strategy was introduced to manage environmental complexity.
• Dijkstra's shortest path search was integrated to refine cost estimation and direction selection.
• Quadratic programming with custom constraints was employed to smooth discrete path points.
• S-curve profiles were utilized for speed planning on the smoothed trajectories.

Main Results:

• The enhanced hybrid A-star algorithm demonstrated significantly improved planning efficiency in simulations and experiments.
• The resulting trajectories exhibited continuous and smooth transitions in heading angle and speed.
• The method effectively handled complex and constrained environments.
• Overall comfort for autonomous vehicles was notably improved.

Conclusions:

• The proposed trajectory planning method offers a substantial improvement over standard hybrid A-star.
• The integration of adaptive expansion, Dijkstra's search, and quadratic programming yields efficient and smooth trajectories.
• This approach enhances the viability of autonomous vehicles in complex, real-world driving conditions.