A Hyper-heuristic for Dynamic Scheduling of Cyber-Physical Production Systems Using Incremental Learning

To ensure efficient dynamic manufacturing scheduling within a Cyber-Physical Production System (CPPS), it is essential to develop reactive control architectures. Heuristic-based optimization algorithms can provide this necessary reactivity and agility. In this paper, a hyper-heuristic is proposed. A set of atomic rules for resource selection are combined in a decisional strategy previously developed. The latter results from an optimization-simulation process. A new incremental learning mechanism is introduced in this article; it allows the system to evolve smoothly to integrate new events from the CPPS and its environment. A comparative study with a metaheuristic and heuristics on 56 instances, with family-dependent setup and processing times, demonstrates the interest of the proposed approach.