Abstract: Conventional ant colony optimization algorithm cannot effectively solve the systems whose optimization performance indices are difficult to be quantifiable. In order to overcome this weakness, a novel Hierarchical Interactive Ant Colony Optimization（HIACO） that the objective function values of the potential solutions are determined by subjective human evaluation is proposed. The structure of a primal Interactive Ant Colony Optimization（IACO） model is designed. Appropriate pheromone update rule and the characters of pheromone in IACO are presented. The ideal of hierarchy, the chance to hierarchy and the method of hierarchy are given. The evaluation way of user is so simple that he or she only needs selecting a mostly interesting individual of current generation and not evaluating quantization of every solution. So user fatigue is reduced efficiently. IACO and HIACO are applied to car styling design. The experimental results demonstrate that the proposed algorithm has good performance.

Key words: ant colony optimization, human-computer interaction, hierarchical, car styling

摘要： 传统蚁群优化算法在求解优化性能指标难以数量化的定性系统问题时无能为力，为此提出一种利用人对问题解进行评价的分层交互式蚁群优化算法。设计了一个基本交互式蚁群优化模型结构，讨论了信息素的更新策略和性质。给出分层的思想、分层的时机和分层的具体实现方法。算法用户参与评价时，只需指出每一代中最感兴趣的解，而不必给出每个解的具体数量值，可以极大降低用户评价疲劳。将算法应用于汽车造型设计，实验结果表明所提出算法具有较高运行性能。

关键词: 蚁群优化, 人机交互, 分层, 汽车造型