In multi-agent systems (MAS), the coalition formation (CF) is an important problem focusing on allocating agents to different tasks. In this paper, three specific CF problems are considered, including the single-task single-coalition formation, the multi-task single-coalition formation, and the multi-task multi-coalition formation. The mathematical models of these three specific problems are formulated with the objective of minimizing the total cost while satisfying the ability requirement constraint. An efficient genetic algorithm with heuristic initialization and repair strategy (GAHIR) is proposed to solve the CF problem. Multiple initialization and repair methods, which utilize the prior knowledge of the specific problems, are proposed to improve the solution quality. Then, these methods are tested to prove their effectiveness. Finally, a comparison experiment about the proposed algorithm against several advanced algorithms is constructed. The results of statistical analysis by the Wilcoxon rank-sum test demonstrate that the proposed GAHIR can obtain better coalition schemes than its competitors in solving the CF problems. Furthermore, GAHIR has faster convergence speed in most instances.

在多代理系统（MAS）中，联盟形成（CF）是一个重要问题，着重于将代理分配给不同任务。本文考虑了三个具体的CF问题，包括单任务单联盟编队，多任务单联盟编队和多任务多联盟编队。制定这三个特定问题的数学模型，旨在在满足能力要求约束的同时将总成本降至最低。提出了一种具有启发式初始化和修复策略（GAHIR）的高效遗传算法来解决CF问题。提出了利用特定问题的先验知识的多种初始化和修复方法，以提高解决方案的质量。然后，对这些方法进行测试以证明其有效性。最后，建立了该算法与几种高级算法的对比实验。Wilcoxon秩和检验的统计分析结果表明，所提出的GAHIR在解决CF问题方面比其竞争对手可以获得更好的联合方案。此外，GAHIR在大多数情况下具有更快的收敛速度。