A Task-Oriented Heuristic for Repairing Infeasible Solutions to Overlapping Coalition Structure Generation
针对资源受限且子可加任务领域中的重叠联盟形成问题，提出一种通用任务导向启发式方法，用于修复不可行解以解决资源冲突，实验表明该方法在资源竞争激烈环境下高效有效。
Overlapping coalition formation (OCF), which provides a natural framework for modeling scenarios where each agent can join and allocate their resources to several completely different coalitions at the same time, has become a very active topic in multiagent systems. For OCF in resource-constrained and subadditive task oriented domains, an agent may not possess sufficient resources to meet the needs of multiple coalitions simultaneously. As a result, there may exist many potential resource conflicts among the rival overlapping coalitions. To tackle such situations, we first present a natural variation of the traditional OCF model and analyze the size of the solution space and the computational complexity of the overlapping coalition structure generation (OCSG) problem. Next, we develop a generic task-oriented heuristic (TOH) for individual repairs that can be used in binary meta-heuristic algorithms to generate overlapping coalitions in a parallel manner. Moreover, we show how the proposed TOH repairs a 2-D individual to resolve resource conflicts and discuss several basic properties. Finally, to evaluate the effectiveness of TOH, we compare it with the existing agent-oriented heuristic for the OCSG problem. The empirical results demonstrate that TOH is of high efficiency and effectiveness in harsh environments with fierce competition over scarce resources.