Deep reinforcement learning for production scheduling

摘要

A method and apparatus are provided for scheduling production in a production facility. A model of a production facility that uses one or more input materials to produce products that satisfy product requests may be determined. Each product request may specify a requested product that may be used at the requested time. Policy and value neural networks may be determined for the production facility. The policy neural network may represent the production behavior scheduled in the production facility, and the value neural network may represent the profit of the product produced in the production facility. Policy and value neural networks may use a model of a production facility during training to create a schedule of production actions in the production facility that satisfies product requests over a specific time period and is associated with a penalty for delaying production of the requested product. .