PPO-Based Dynamic Task Scheduling for Automated Filter Rod Warehouse Systems

In the production plan-driven manufacturing mode, the filter rod warehouse often encounters various uncertainties in actual operation, such as fluctuating demand between multiple batches, differences in curing times for different filter rod types, and dynamic changes in environmental conditions. Traditional methods fail to achieve high operational efficiency in loading/unloading tasks under such variable conditions. To address this issue, this study proposes a batch loading/unloading optimization model for filter rod warehouses based on the Proximal Policy Optimization algorithm. The model features a dual-action decision mechanism: first, the intelligent agent selects optimal storage locations for incoming materials based on current shelf status; second, it determines corresponding unloading tasks to achieve dynamic combination and sequencing of loading/unloading operations. Additionally, the model incorporates an adaptive weighted distance strategy to guide the agent in learning optimal matching relationships between loading and unloading tasks through high-frequency interactions. Simulation results demonstrate approximately 13.5% improvement in operational efficiency with good generalization capability.