Multi-AUV Kinematic Task Assignment Based on Self-Organizing Map Neural Network and Dubins Path Generator.

Area of Science:

• Robotics
• Artificial Intelligence
• Control Systems

Background:

• Multi-AUV systems face challenges in task assignment due to kinematic constraints.
• Underactuated AUVs require sophisticated algorithms for efficient operation.
• Trajectory planning failures can occur due to obstacles or vehicle limitations.

Purpose of the Study:

• To propose an improved task assignment algorithm for multi-AUV systems.
• To address kinematic constraints and potential trajectory planning failures.
• To ensure successful task completion for AUVs in complex environments.

Main Methods:

• Combines an improved Self-Organizing Map (SOM) neural network with the Dubins Path algorithm.
• Utilizes workload balance and neighborhood functions for initial task assignment.
• Implements dynamic task re-assignment by adjusting SOM neural weights when kinematic constraints or obstacles impede path planning.

Main Results:

• The algorithm effectively assigns tasks to AUVs, considering workload balance.
• Dynamic re-assignment ensures path feasibility under kinematic constraints and obstacles.
• Simulation results demonstrate successful path planning and task accomplishment for multi-AUV systems.

Conclusions:

• The proposed algorithm enhances task assignment efficiency for multi-AUV systems.
• It provides a robust solution for AUVs operating under kinematic constraints.
• The integration of SOM and Dubins Path algorithms offers a viable approach for complex AUV missions.