A Dynamic Bioinspired Neural Network Based Real-Time Path Planning Method for Autonomous Underwater Vehicles.

Area of Science:

• Robotics
• Artificial Intelligence
• Marine Engineering

Background:

• Real-time path planning for autonomous underwater vehicles (AUVs) in 3D unknown environments is challenging.
• Bioinspired neural networks (BINNs) offer advantages like no learning requirement but face issues with large environments and obstacle sizes.
• Existing BINNs struggle with computational complexity and repeated paths for AUVs.

Purpose of the Study:

• To propose an improved dynamic bioinspired neural network (BINN) for enhanced real-time path planning in 3D unknown underwater environments.
• To address the computational complexity and repeated path problems associated with traditional BINNs for AUVs.
• To enhance the efficiency and effectiveness of AUV navigation in complex underwater scenarios.

Main Methods:

• An improved dynamic BINN is proposed, treating the AUV as the core and sizing the network based on sensor detection range.
• The BINN moves with the AUV to reduce computational load.
• A virtual target and target attractor concepts are introduced to ensure effective navigation and improve neural activity efficiency.

Main Results:

• The proposed dynamic BINN method effectively reduces computational complexity for AUV path planning.
• The virtual target mechanism allows automatic avoidance of large obstacles.
• Experiments in various 3D underwater environments demonstrate the method's efficiency.

Conclusions:

• The improved dynamic BINN provides an efficient solution for real-time path planning for AUVs in complex 3D environments.
• The method overcomes limitations of traditional BINNs, offering better performance and reduced computational demands.
• This approach enhances AUV autonomy and navigation capabilities in challenging underwater settings.