Data-driven scheduling for smart shop floor via reinforcement learning with model-based clustering algorithm

Abstract

Various information technologies provide the manufacturing system massive data, which can support the decision optimization of product lifecycle management. However, the lack of effective use for advanced technologies prevents manufacturing industry from being automated and intelligent. Therefore, this paper proposes the smart shop floor and implementation mechanism. Meanwhile, based on the clustering and reinforcement learning, the brain agent and scheduling agent are designed to determine the approriate rule according to the shop floor state information, thus realizing the data-driven real-time scheduling. Experimental results show that the smart shop floor can effectively deal with disturbance events and has better performance compared with composite dispatching rules.