Abstract: To address the problems of imbalanced resource allocation and difficulty in collaborative multi-objective optimization for multi-Unmanned Aerial Vehicle (UAV) task assignment in large-scale unbalanced scenarios, an integrated hybrid centralized-distributed assignment framework that combines improved clustering and coalition formation game is proposed. First, the optimization objectives including resource utilization, total path length and energy consumption are modeled as a utility function. Subsequently, the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm is improved to achieve adaptive adjustment of neighborhood parameters through inflection point identification of the k-distance graph, and hierarchical processing of noise points is implemented via near-cluster matching, so as to generate spatially compact task clusters and reduce the scale of the problem. On this basis, a greedy coalition formation game model is constructed based on the obtained task clusters. The initialization mechanism constrained by resource matching and endurance range, as well as the multi-strategy decision-making of UAVs are introduced into the model, which takes the utility function as the quantitative basis and the utilitarian order as the decision criterion to guide the system to converge to the Nash equilibrium, thus realizing dynamic and balanced resource allocation. Finally, simulation results demonstrate that the comprehensive performance of the proposed framework on core optimization objectives is significantly superior to that of the benchmark comparison algorithms. It provides an efficient technical approach that balances comprehensive multi-objective optimization performance and stability for multi-UAV cooperative task assignment in unbalanced scenarios.

Key words: unmanned aerial vehicle (UAV), task assignment, Density-based spatial clustering of applications with noise (DBSCAN), greedy coalition formation game, multi-objective optimization, resource utilization rate