A Path-Planning Scheme for Autonomous Vehicle Navigation Integrating BJA* and Improved DWA Algorithms.

Area of Science:

• Robotics
• Artificial Intelligence
• Autonomous Systems

Background:

• Traditional A* and Dynamic Window Approach (DWA) algorithms have limitations in dynamic environments, path smoothness, and real-time obstacle avoidance.
• A* struggles with dynamic environments and path quality, while DWA can get stuck in local optima and requires extensive parameter tuning.

Purpose of the Study:

• To propose a novel autonomous driving vehicle navigation path-planning scheme by fusing Bidirectional Jump point A* (BJA*) and an improved DWA.
• To enhance global path planning efficiency and local obstacle avoidance robustness for autonomous vehicles.

Main Methods:

• The proposed scheme fuses BJA* with an improved DWA, enhancing A*'s global efficiency using a 24-neighborhood search and bidirectional jump-point strategy.
• DWA's local robustness is improved by optimizing the evaluation function and integrating global path information.
• MATLAB (2022b) simulations were used for comparative analysis.

Main Results:

• BJA* demonstrated significant improvements in path length, traversal time, and turn number compared to improved A* methods.
• The integrated local obstacle avoidance resulted in more stable speed control for the vehicle in dynamic scenarios.
• Ablation experiments confirmed the contribution of specific improvements in BJA*.

Conclusions:

• The fused BJA*-DWA navigation scheme effectively addresses the limitations of traditional algorithms for autonomous driving.
• The proposed method offers enhanced global path planning and robust local obstacle avoidance, leading to more stable and efficient vehicle navigation.