论文部分内容阅读
针对Flow-shop排序问题的固有复杂性,设计了一种遗传强化学习算法.首先,引入状态变量和行动变量,把组合优化的排序问题转换成序贯决策问题加以解决;其次,设计了一个Q-学习算法和基于组合算子的遗传算法相集成,遗传算法利用染色体的优良模式及其适应值信息来指导智能体的学习过程,提高学习效率和效果,强化学习则对染色体进行局部优化进而改良遗传群体,二者有机结合共同解决Flow-shop排序问题;再次,提出了多种适应性策略,使算法关键参数能够周期性递变,以更好地在深度搜索和广度搜索之间均衡;最后,仿真优化实验结果验证了该算法的有效性.
Aiming at the inherent complexity of Flow-shop scheduling problem, a genetic reinforcement learning algorithm is designed.Firstly, by introducing state variables and action variables, the combinatorial optimization ordering problem is converted into a sequential decision problem to be solved.Secondly, a Q - Learning algorithm and compositionalgorithm based on the integration of genetic algorithms, genetic algorithms using the excellent model of chromosomes and fitness information to guide the learning process of the agent to improve the learning efficiency and effectiveness, strengthen the learning of the chromosomes and then optimize the local optimization Genetic groups, the two are combined to solve the Flow-shop scheduling problem. Thirdly, a variety of adaptive strategies are proposed to make the key parameters of the algorithm periodically change to better balance between depth search and breadth search. Finally, Simulation results show that the proposed algorithm is effective.