Reinforcement learning-based multiobjective and multiconstraint production scheduling for precast concrete

Optimizing the production scheduling for precast concrete can significantly reduce lead times and enhance delivery efficiency. The performance of traditional scheduling models for precast concrete can be enhanced due to the unique characteristics of the concrete during production. However, the efficiency of these models decreases as problem complexity and uncertainty increase. This study introduces practical considerations, such as a limited number of molds, buffers, uncertainty of order arrivals, and vehicles. Furthermore, a multiobjective optimization scheduling model is developed to address the requirements of modern industrial development, by considering the on-time delivery rate, total processing time, and workstation utilization rate. A reinforcement learning algorithm-based solution is devised and validated through real-world case studies. This methodology effectively addresses the challenges of production scheduling for precast concrete in a multiconstraint, multiobjective real-world scenario with uncertain order arrival times. By adopting this approach, small and medium-sized precast manufacturers can enhance their responsiveness to unpredictable scheduling issues, thereby significantly improving the efficiency of precast concrete production.