Abstract: To achieve cooperative optimization of a genetic programming (GP)-based dynamic system between structure and parameter during the evolutionary design process, the diversity preservation strategy of evolutionary design topology population is discussed and a topology fitness-sharing & crowding cooperative search algorithm is proposed. This algorithm avoids calculate niche’s radius and punish the topology subgroup by self-adaptive fitness-function, thus the diversity of the topology is well preserved and convergence is prohibited. A case is employed to demonstrate the practicality of applying balanced topology and parameter search in evolutionary design of the dynamic system.

Key words: genetic programming, bond graphs, evolutionary design, topology search

摘要： 为了解决基于遗传编程(GP)的动态系统进化设计过程中拓扑和参数协同优化的问题，讨论了基于GP的进化设计种群拓扑多样性保存策略，提出了一种拓扑适应值共享-拥挤协同搜索算法。该算法避免计算小生境半径、通过自适应适应度函数来惩罚拓扑子群，保证了拓扑多样性和阻止局部收敛的发生。实验结果表明，该算法保证了动态系统进化设计中拓扑和参数同步搜索的平衡，有效地克服了局部收敛，能确保获得理想的设计结果。

关键词: 遗传编程, 功率键合图, 进化设计, 拓扑搜索

CLC Number: 

• TP301.6