Motion planning for autonomous vehicle based on radial basis function neural network in unstructured environment.

Area of Science:

• Robotics
• Artificial Intelligence
• Control Systems

Background:

• Autonomous vehicles face significant challenges navigating unstructured environments due to irregular road shapes and real-time planning demands.
• Nonholonomic constraints inherent to vehicles further complicate motion planning in complex terrains.

Purpose of the Study:

• To develop and validate a robust motion planning method for autonomous vehicles operating in unstructured environments.
• To enhance path flexibility, smoothness, and safety compared to existing methods.

Main Methods:

• A motion planning method utilizing a Radial Basis Function (RBF) neural network is proposed.
• The algorithm identifies drivable regions from perception grid maps and uses gradient descent to train the RBF network.
• Sample points are randomly selected within the drivable region for network training.

Main Results:

• The proposed RBF-based method generates flexible, smooth, and safe paths adaptable to various road shapes.
• Simulations and experiments confirm the effectiveness of the algorithm in unstructured environments.
• Comparison with the Rapidly-exploring Random Tree (RRT) method shows superior performance in path planning quality.

Conclusions:

• The RBF neural network-based motion planning method is highly effective for autonomous vehicles in unstructured environments.
• The approach offers improved motion quality and adaptability compared to traditional methods like RRT.