A Multi-action Reinforcement Learning Algorithm for Energy-efficiency Blocking Flow-shop Scheduling Problem

With the increasingly serious ecological problems, energy-efficient scheduling, an effective approach to achieve sustainable development and green manufacturing, has attracted much attention by taking both economic effect and energy conservation into account. This paper addresses an energy-efficient scheduling of the distributed blocking flow-shop problem (EDBFSP) to minimize both makespan and total energy consumption. The mixed-integer linear programming (MILP) model of EDBFSP is designed. A multi-action reinforcement learning algorithm based on problem-specific knowledge called multi-greedy policy optimization (multi-GPO) is proposed to solve the EDBFSP. In addition, after analyzing the characteristics of the problem, an energy-saving strategy and an acceleration strategy are designed to further optimize the solution. Experiments in a large number of benchmark tests have testified that the multi-GPO is superior to the state-of-the-art algorithms in terms of efficiency and importance in solving EDBFSP.