We study the Dynamic Flexible Job Shop Scheduling Problem (DFJSSP), which is a well-known combinatorial optimization problem where new jobs arrive dynamically over time and have to be scheduled on a sequence of machines. Unlike most works that look at makespan or related criteria, we study a new cost minimization objective for the DFJSSP. We propose a new deep reinforcement learning method for scheduling and show that our method outperforms benchmark methods.
Deep Reinforcement Learning for the Dynamic Flexible Job Shop Scheduling Problem with Cost Optimization
78 views
5 Downloads