Multifactorial PSO-FA Hybrid Algorithm for Multiple Car Design Benchmark
Multifactorial PSO-FA Hybrid Algorithm for Multiple Car Design Benchmark
复制标题
DOI：
10.1109/smc.2019.8914649
复制
发表时间：
2019-10
期刊：
影响因子：
--
通讯作者：
Heng Xiao;Gen Yokoya;T. Hatanaka
中科院分区：
文献类型：
--
作者：
Heng Xiao;Gen Yokoya;T. Hatanaka
进化多任务作为进化计算的一种新应用，目前受到人们的关注，其中搜索群体并行处理独立的问题。进化多因子优化是近年来提出的一种进化多任务优化方法，它是受人类处理多任务的认知能力的启发而提出的。实际的进化多因子优化使用不同问题之间的关系，并为每个问题分配资源。基于群体搜索的并行性有望成为多任务优化问题的一个很好的解决方案，其中需要对不同的任务进行优化。本文提出了一种基于混合群智能的多因子优化算法。将粒子群模型与萤火虫模型相结合，利用多因子进化算法中的技能因子进行多因子分解。然后，我们将该算法应用于多车结构设计基准，其中三辆不同的汽车具有较轻但令人满意的刚度和安全性。换句话说，同时三辆车的设计问题被视为一个多任务优化问题。实验结果表明，该算法比基于群智能的多因子优化算法具有更高的搜索效率。
Evolutionary multitasking is currently under focus as a new application of evolutionary computation where a search population operates independent problems in parallel. An evolutionary multifactorial optimization has proposed in recent years as an instance of evolutionary multitasking, it is inspired by human cognition ability processing several tasks. An actual evolutionary multifactorial optimization uses a relationship among different problems and assigns resources to each problem. The parallelism of population-based search is expected to be a good solver for the multitask optimization where different tasks should be optimized. In this paper, we propose a hybrid swarm intelligence based multifactorial optimization. A hybrid swarm is combined with particle swarm model and firefly model, and multi-factorization by using the skill factor that is used in the multifactorial evolutionary algorithm. Then we apply the proposed algorithm to the multiple car structure design benchmark where three different cars are treated to have a light but a satisfying stiffness and safety. In other words, the simultaneous three cars design problem is treated as a multitask optimization. The experimental results show that the hybridization gives more efficient search performance than the base swarm intelligence based multifactorial optimization.