A Metaheuristic Optimization Algorithm for Task Clustering in Collaborative Multi-Cluster Systems.

Area of Science:

• Robotics and Artificial Intelligence
• Operations Research
• Computational Science

Background:

• Unmanned Aerial Vehicle (UAV) swarm missions require efficient task grouping in complex 3D environments.
• Existing methods often suffer from spatial information loss and premature convergence in high-dimensional task allocation.

Purpose of the Study:

• To develop an advanced clustering method for optimizing air-ground integrated UAV swarm task allocation.
• To improve the diversity, adaptability, and solution quality of task grouping in 3D environments.

Main Methods:

• Proposed a 3D spatial task data preprocessing technique and a golden spiral distribution-based hybrid initialization strategy.
• Developed a Dual-Prototype Metaheuristic K-Means (DPM-Kmeans) algorithm with dual-modal prototypes (row and column) for simultaneous global and local search.
• Implemented a collaborative multi-constraint, dynamically weighted optimization model integrating task requirements and flight distance.

Main Results:

• DPM-Kmeans demonstrated a 2-10% improvement in Sum of Squared Errors (SSE), Silhouette Coefficient (SC), and Davies-Bouldin Index (DB) compared to traditional K-means and other meta-heuristic algorithms.
• The method exhibited superior convergence speed and solution quality.
• Achieved excellent scalability and robustness in large-scale, multi-constraint 3D scenarios.

Conclusions:

• The proposed DPM-Kmeans algorithm effectively addresses the task-grouping problem for UAV swarms in 3D environments.
• The dual-modal prototype framework and hybrid initialization enhance search capabilities and prevent premature convergence.
• DPM-Kmeans offers a robust and scalable solution for complex UAV mission planning.