A self-adaptive grey wolf optimizer incorporating multi-head deep Q-network for reconfigurable job-shop scheduling with multi-skilled workers

In the context of Industry 5.0, manufacturing systems are shifting towards flexible and human-centred production. In practical workshops, machine reconfiguration involves parameter adjustment, tool replacement and system calibration, with efficiency affected by workers’ skill levels and availability. This article studies a reconfigurable job-shop scheduling problem with multi-skilled workers for machine reconfiguration (RJSP-MWR). First, a mathematical model of the RJSP-MWR is developed, incorporating workers’ skill levels and availability into reconfiguration time modelling, with the objectives of minimizing the makespan and reconfiguration cost. Then, a self-adaptive grey wolf optimizer incorporating multi-head deep Q-network (MHDQN-GWO) is proposed. The multi-head deep Q-network adaptively adjusts the parameters and search operators of the grey wolf optimizer. Finally, comparative experiments on 18 instances demonstrate that MHDQN-GWO exhibits superior convergence and stability compared with other algorithms, achieving the best solutions on 94.4% of instances. The method is successfully applied to a real-world ultra-high-speed optical module production workshop.