AI 中文总结
针对高度约束的多目标决策问题，介绍IMAP-IGS框架，将偏好聚合嵌入进化适应度函数，给出两种实现。实验表明，在利益相关者偏好冲突时，该框架优于其他方法，优势体现在解决高度约束决策环境中的偏好冲突。
AI 中文摘要
高度约束的多目标设计和决策问题因严格的可行性要求和利益相关者偏好冲突而难以解决。进化算法虽广泛用于此类问题，但通常将约束处理与偏好优化分开。基于帕累托的方法提供一组权衡解决方案，但需后处理以选择偏好选项，而标量化可能扭曲不同利益相关者的偏好。本文介绍了集成偏好最大化（IMAP）及其代际求解器（IMAP-IGS），它们将偏好聚合直接嵌入进化适应度函数。约束违反偏好函数引导搜索趋向可行性，最终得分仅反映利益相关者偏好。该框架可集成到任何使用标量适应度评估的进化算法中。给出了IMAP-BRKGA用于组合优化和IMAP-GA-II用于连续优化这两种实现。在DAS-CMOP、MO-VRPTW和HVASP基准测试表明，当利益相关者偏好冲突时，IMAP始终优于加权和及基于帕累托的方法。当偏好一致时，这些优势基本消失，表明IMAP的优势在于解决高度约束决策环境中的偏好冲突。
英文摘要
Highly constrained multi-objective design and decision problems are difficult to solve because of strict feasibility requirements and conflicting stakeholder preferences. Evolutionary algorithms are widely used for these problems but typically separate constraint handling from preference optimisation. Pareto-based methods provide a set of trade-off solutions but require post-processing to select a preferred option, while scalarisation can distort heterogeneous stakeholder preferences. This paper introduces the Integrative Maximisation of Aggregated Preferences (IMAP) and its Inter-Generational Solver (IMAP-IGS), which embed preference aggregation directly into the evolutionary fitness function. A constraint-violation preference function guides the search toward feasibility, while final scores reflect only stakeholder preferences. The framework can be integrated into any evolutionary algorithm that uses scalar fitness evaluation. Two implementations are presented: IMAP-BRKGA for combinatorial optimisation and IMAP-GA-II for continuous optimisation. Testing on DAS-CMOP, MO-VRPTW, and HVASP benchmarks shows that IMAP consistently outperforms weighted-sum and Pareto-based approaches when stakeholder preferences conflict. These advantages largely disappear when preferences align, indicating that IMAP's strength lies in resolving preference conflicts in highly constrained decision environments.
