Research on the Coalition Formation of Multi-Agent Systems for Forest Fire Prevention and Control Tasks

The multi-agent systems (MAS) for forest fire prevention and control are large-scale, with varied structural characteristics, leading to high complexity in management. This paper delves into the Coalition Formation (CF) strategy of MAS in this scenario. Considering the characteristic and the actual utility of coalitions, as well as the multiple dimensions of constraints and optimization indicators, a model for the agent CF problem is established. A two-stage hybrid intelligent algorithm is proposed to solve the problem. The algorithm is based on the K-means Clustering algorithm, Variable Neighborhood Search (VNS) algorithm and Genetic Algorithm (GA). This method can form a high-quality feasible CF scheme, which provides support for the intelligent control of fire system resources and the implementation of subsequent task allocation.