Research on a hybrid neural network task assignment algorithm for solving multi-constraint heterogeneous autonomous...

Area of Science:

• Robotics
• Artificial Intelligence
• Operations Research

Background:

• Underwater exploration relies on multi-robot task assignment, crucial for military, fishery, and energy sectors.
• Existing heuristic methods struggle with optimal task assignment in heterogeneous autonomous underwater vehicle (AUV) clusters, especially with autonomous decision-making.
• The complexity of multi-constrained detection task assignment for AUV clusters necessitates advanced algorithmic solutions.

Purpose of the Study:

• To propose a novel Fast Graph Pointer Network (FGPN) method for efficient and accurate multi-constrained task assignment in heterogeneous AUV clusters.
• To enhance the efficiency of task assignment for detection/communication AUV clusters while ensuring solution accuracy.
• To address the challenge of large-scale cooperative task assignment by transforming it into a Multiple Traveling Salesman Problem (MTSP).

Main Methods:

• A two-stage detection algorithm involving task node clustering based on communication distance.
• A neural network model utilizing Graph Pointer Network (GPN) for local task assignment post-clustering.
• Integration of GPN with a genetic algorithm to form the proposed Fast Graph Pointer Network (FGPN) method.
• Formulation of a large-scale cluster cooperative task assignment problem and a detection/communication cooperative work mode.

Main Results:

• The FGPN method demonstrates superior solution efficiency compared to traditional heuristic methods for task assignment scales of 300 to 2,000 task nodes.
• The solution quality achieved by FGPN is comparable to that of heuristic methods.
• Experimental validation confirms the effectiveness of the proposed FGPN algorithm for large-scale cooperative task assignment problems.

Conclusions:

• The FGPN method offers an effective approach to solving complex, large-scale task assignment problems for heterogeneous AUV clusters.
• The proposed method provides a valuable reference for addressing similar large-scale task assignment challenges in diverse fields.
• This research advances autonomous decision-making capabilities in multi-robot systems for underwater applications.