July 1, 2022
Reinforcement Learning Approach To Scheduling Of Precast Concrete Production
Kim, Taehoon; Kim, Yong-woo; Lee, Dongmin; Kim, Minju. (2022). Reinforcement Learning Approach To Scheduling Of Precast Concrete Production. Journal Of Cleaner Production, 336.
Abstract
The production scheduling of precast concrete (PC) is essential for successfully completing PC construction projects. The dispatching rules, widely used in practice, have the limitation that the best rule differs according to the shop conditions. In addition, mathematical programming and the metaheuristic approach, which would improve performance, entail more computational time with increasing problem size, let alone its models being revised as the problem size changes. This study proposes a PC production scheduling model based on a reinforcement learning approach, which has the advantages of a general capacity to solve various problem conditions with fast computation time and good performance in real-time. The experimental study shows that the proposed model outperformed other methods by 4-12% of the total tardiness and showed an average winning rate of 77.0%. The proposed model could contribute to the successful completion of off-site construction projects by supporting the stable progress of PC construction.
Keywords
Precast Concrete; Reinforcement Learning; Deep Q -network; Production Scheduling; Minimize; Model
