Deep Reinforcement Learning Based Intelligent Job Batching in Industrial Internet of Things

The ever-developing Industrial Internet of Things (IIoT) is promoting the transformation of traditional manufacturers to intelligent manufacturing. Intelligent job batching, as one of the important parts of intelligent production in IIoT, is desired to group jobs with similar features into one batch under the constraint of batch capacity while considering the manufacturing target. This work formulates the job batching problem as a Markov Decision Process and proposes a deep reinforcement learning (DRL) based method to achieve intelligent job batching. The job batching model is based on the pointer network. The convergence of the model under different parameters is analyzed, and the performance of the method is evaluated by comparing it with the manual result, K-means algorithm, and multiple meta-heuristic algorithms via real production data. Experiments show that the proposed method can produce better solution in terms of the feature difference within a batch and the total batch number, especially in large-scale manufacturing scenarios.