Local Path Planning of Autonomous Vehicle Based on an Improved Heuristic Bi-RRT Algorithm in Dynamic Obstacle...

Area of Science:

• Robotics
• Artificial Intelligence
• Autonomous Systems

Background:

• Existing Rapidly-exploring Random Trees (RRT) algorithms face challenges in local path planning for autonomous vehicles, particularly with path coherence between front and rear frames.
• Autonomous vehicles require robust obstacle avoidance in unknown dynamic environments, a task where current RRT variants are insufficient.

Purpose of the Study:

• To propose an improved heuristic Bi-RRT algorithm for effective local path planning and obstacle avoidance in dynamic environments.
• To address the path coherence problem in autonomous vehicle navigation.

Main Methods:

• Introduced vehicle constraints considering driver habits and an obstacle-free direct connection for two random trees.
• Implemented multi-sampling biased towards the target state and parent node selection using a comprehensive measurement index.
• Utilized adaptive greedy step size with target direction and path reorganization for curvature continuity and coherence.

Main Results:

• The proposed Bi-RRT algorithm demonstrated efficient performance in simulations, generating smoother paths in shorter times compared to four other algorithms.
• Experiments in dynamic environments confirmed the algorithm's ability to produce differentiable, coherent paths.
• The algorithm effectively minimizes redundant path points and ensures smooth transitions between path segments.

Conclusions:

• The improved heuristic Bi-RRT algorithm is suitable for autonomous vehicle local path planning in unknown dynamic environments.
• The algorithm enhances path smoothness, coherence, and efficiency, leading to improved ride comfort and vehicle stability.
• This approach offers a viable solution for complex navigation challenges faced by autonomous vehicles.