Optimization Control Of Multiple CSPS With Dynamic Pickup Point

In modern manufacturing industry,Conveyor Served Production Station(CSPS)is a kind of production line system composed of processing stations,which can unload and process workpieces.In the actual production process,the production line with only one station can not meet large-scale production.Therefore,there are several processing stations on the conveyor belt,which is called the Multiple CSPS.To improve the processing efficiency of the production line and reduce the processing cost,it is necessary to study the optimal control of this system and select the control strategy rationally.Compared with fixed point pickup used in traditional CSPS,the dynamic pickup station can unload the workpiece directly in the working area of the mechanical arm without waiting,the production cycle of the system is shorter.This dissertation studies a multiple CSPS with dynamic pickup point.The decision of upstream station affects the operation of downstream station,so it is difficult to establish Markov model to solve the optimal strategy in theory.Therefore,This dissertation first establishes a reinforcement learning model for the system.The remaining amount of the buffer and the state of the manipulator are taken as the joint state variate of the station,the look-ahead distance is taken as the control variate.Then,Wolf-PHC algorithm is used to solve the optimal strategy of each station.The effects of delay waiting,changing system parameters and the cost of the end station on the system performance are analyzed by simulation.The results show that algorithm can effectively improve the processing rate of the system.In addition,this dissertation studies the application of centralized control in the system.The decision time of the system is set to the moment when processing or unloading operation of any station is completed,and the system is optimized using DQN algorithm.The experimental results show that DQN algorithm can improve the system processing rate and reduce the average cost of the system.In terms of improving the Equilibrium,Wolf-PHC is superior to DQN.