应用改进Q-learning算法解决柔性作业车间调度问题

Application of improved Q-learning algorithm to solve flexible job shop scheduling problem

柔性作业车间调度是目前加工系统中的一个重要调度问题,对于该问题的解决方法大都存在速度较慢等缺陷,提出一种改进的Q-learning算法,设计了动态的状态空间及动作集,通过引入“Sigmoid”函数作为动态选择策略改进Q-learning算法,使改进后的算法前期随机选择动作,后期在随机选择动作和选择奖励值最高的动作中动态变化,有效改善了传统Q-learning算法容易陷入局部最优且收敛速度慢等缺陷。将改进Q-learning算法应用到TSP问题中,证实改进算法的普适性和可行性,再将其应用解决柔性调度问题中,证实了其改进的有效性,提升了解决柔性作业车间调度问题的速度和精度。

Flexible job shop scheduling is an important scheduling problem in the current processing system.Most of the solutions to this problem have problems such as slow speed.This paper proposes an improved Q-learning algorithm and designs a dynamic state space and action set.Improve the Q-learning algorithm by introducing the"Sigmoid"function as a dynamic selection strategy,so that the improved algorithm randomly selects actions in the early stage,and dynamically changes in the later stage of the random selection of actions and the action with the highest reward value,effectively improving the traditional Q-learning algorithm It is easy to fall into defects such as local optimum and slow convergence speed.This article first applies the improved Q-learning algorithm to the TSP problem,confirms the universality and feasibility of the improved algorithm,and then applies it to solve the flexible scheduling problem,confirms the effectiveness of its improvement,and improves the solution to flexible job shop scheduling.The speed and accuracy of the problem.