Cloud Manufacturing Task Scheduling Under Machinery Breakdown Based on Deep Reinforcement Learning

Abstract

Task scheduling is one of the major research issues in cloud manufacturing. At present, most studies choose heuristic algorithms to solve task scheduling problems. However, cloud manufacturing environment is highly complex, and dynamic events can easily affect the execution of scheduling scheme. Heuristic algorithms are difficult to satisfy requirements of dynamic scheduling. In recent years, deep reinforcement learning has been applied to many fields. In this paper, a deep reinforcement learning algorithm named as Long Short Term Memory-Double Deep Q-Network (LSTM-DoubleDQN) is designed to solve the scheduling problem in cloud manufacturing, and the rescheduling method under machinery breakdown is also explored. The results show that the convergence speed of LSTM-DoubleDQN is faster than Deep Q-Network (DQN), Long Short Term Memory-Deep Q-Network (LSTM-DQN) and Double Deep Q-network (DoubleDQN). The QoS of scheduling scheme of LSTM-DoubleDQN is better, which is more suitable for solving the scheduling problem in cloud manufacturing. The effectiveness of rescheduling method is also verified.