基于蛙跳算法的分布式装配混合流水车间调度

Distributed assembly hybrid flow shop scheduling based on shuffled frog leaping algorithm with Q-learning

为了降低延迟交货率,针对以总延迟时间为优化目标的分布式装配混合流水车间调度问题,提出基于Q-学习的蛙跳算法.设计了问题的三串编码方法,并给出解码过程.将Q-学习嵌入到蛙跳算法的模因组搜索过程中,Q-学习算法包括由全局搜索、邻域搜索和解的接收准则组成的动作集合,和基于种群精英解和离散度而构建的6种状态.在算法运行过程中,根据种群的状态,利用Q-学习动态地选择执行的模因组搜索策略.实验结果表明:与现有算法相比,基于Q-学习的蛙跳算法在112个实例中均能获得更好或者相同的结果,表明基于Q-学习的蛙跳算法在求解考虑运输和装配的分布式混合流水车间调度问题方面具有较强优势.

In order to reduce the delayed delivery rate,a shuffled frog leaping algorithm with Q-learning(QSFLA)was proposed for distributed assembly hybrid flow shop scheduling problem to minimize total tardiness.A three-string coding method was provided.Q-learning was embedded in the shuffled frog leaping algorithm and applied to the Memeplexes search process.The Q-learning process includes an action set composed of global search,neighborhood search and solution acceptance criteria,and six states based on the elite solution of population and dispersion.In the process of Q-learning,the state of the population was evaluated,and a Memeplex search strategy was chosen and executed by Q-learning according to the state of the population.Extensive experiments were conducted.Discrete shuffled frog leaping algorithm(DSFLA)can obtain the better or the same results compared with the comparing algorithm in 112 instances.The computational results indicate that shuffled frog leaping algorithm with Q-learning has promising advantages on solving distributed assembly hybrid flow shop scheduling problem.