论文部分内容阅读
研究带有柔性站点的多工序传送带给料加工站系统的优化控制问题。根据系统运行特点,文章把系统分为切换控制和Look-ahead协同控制两层决策。首先,对于上层的柔性站点切换控制问题,根据CMAC神经网络收敛速度快,适应能力强的特点,采用基于CMAC神经网络的Q学习算法进行策略优化,解决工序间控制问题;其次,对于下层的Look-ahead协同控制问题,运用Wolf-PHC多Agent学习算法,解决工序内协同问题。实验结果表明,通过引入柔性站点建立两层决策体系,整个系统的工件处理率明显提高,基于CMAC神经网络的Q学习算法收敛速度更快,空间复杂度更低。
Study on Optimal Control of Multi-process Conveyor Belt Feed Station System with Flexible Stations. According to the characteristics of system operation, the article divides the system into two-layer decision-making of switching control and Look-ahead collaborative control. Firstly, based on CMAC neural network’s fast convergence and strong adaptability, the Q-learning algorithm based on CMAC neural network is used to optimize the control strategy of the top flexible site and solve the problem of control between processes. Secondly, -ahead collaborative control problems, the use of Wolf-PHC multi-Agent learning algorithm to solve the intra-process collaboration problems. Experimental results show that by introducing a flexible site to establish a two-tier decision-making system, the processing rate of the whole system is obviously improved. The Q learning algorithm based on CMAC neural network converges faster and has less space complexity.