Flexible job shop real-time scheduling method based on Dueling architecture deep reinforcement learning

The invention discloses a flexible job shop real-time scheduling method based on Dueling architecture deep reinforcement learning, and the method comprises the steps: constructing a DFJSP mathematical model for random arrival of workpieces, and designing a flexible job shop environment state space; the workshop environment state space comprises the total number of workshop processing machine tools, the average number of independently processed workpieces in unit time, the number of newly inserted workpieces, the estimated average utilization rate of the machine tools and the actual average utilization rate of the machine tools, designing a scheduling distribution rule, designing an instant reward function based on a DFJSP mathematical model, designing an action strategy, and designing a scheduling agent. The scheduling agent comprises an online network and a target network, a rescheduling moment is set, the sum of working procedure numbers required by the rescheduling moment is calculated, and the workshop is