Research on Coordinated Scheduling of Cross-Enterprise Production and Transportation Under Shared Manufacturing

To accomplish collaborative production across upstream and downstream enterprises under shared manufacturing, cross-enterprise production and transportation coordination scheduling challenge is explored. A cross-enterprise scheduling model with process sequence constraints is developed with the optimization objectives of job tardiness and overall cost. To solve the problem, this chapter proposes an improved deep Q-network algorithm with a prioritized experience replay mechanism (IDQN-PER). The state is built using real-time data from jobs and machines. Actions are described based on the due date of jobs and the cost of machines. The reward function is formed based on objectives. IDQN-PER employs the Dueling DQN to estimate the state value and action advantage independently, improving the model’s stability. Meanwhile, combining the DDQN successfully eliminates the problem of Q-value overestimation, and adding the prioritized experience replay pool improves the convergence of the algorithm. The experimental findings show that solving the issue using the IDQN-PER produces less tardiness and cost, as well as faster convergence, as compared to the DQN and D3QN.