Distributed flexible job shop scheduling method and device based on deep reinforcement learning

The invention provides a distributed flexible job shop scheduling method and device based on deep reinforcement learning. The scheduling method comprises the following steps: S1, acquiring workpiece information and production environment information; s2, based on the workpiece information and the production environment information, the first intelligent agent allocates workpieces to all the machining units, workpiece allocation information is obtained, and the first intelligent agent configures a trained workpiece allocation strategy network; and S3, based on the workpiece distribution information and the production environment information of the processing unit, the second intelligent agent distributes a processing sequence for the workpieces in the processing unit and distributes a machine for each process of the workpieces, and the second intelligent agent configures a trained operation sorting-machine distribution strategy network. The first intelligent agent and the second intelligent agent form a multi-