A reinforcement learning iterated local search for makespan minimization in additive manufacturing machine scheduling problems

Additive manufacturing - also known as 3D printing - is a manufacturing process that is attracting more and more interest due to high production rates and reduced costs. This paper focuses on the scheduling problem of multiple additive manufacturing machines, recently proposed in the scientific literature. Given its intractability, instances of relevant size of additive manufacturing (AM) machine scheduling problem cannot be solved in reasonable computational times through mathematical models. For this reason, this paper proposes a Reinforcement Learning Iterated Local Search meta-heuristic, based on the implementation of a Q-Learning Variable Neighborhood Search, to provide heuristically good solutions at the cost of low computational expenses. A comprehensive computational study is conducted, comparing the proposed methodology with the results achieved by the CPLEX solver and to the performance of an Evolutionary Algorithm recently proposed for a similar problem, and adapted for the AM machine scheduling problem. Additionally, to explore the trade-off between efficiency and effectiveness more deeply, we present a further set of experiments that test the potential inclusion of a probabilistic stopping rule. The numerical results evidence that the proposed Reinforcement Learning Iterated Local Search is able to obtain statistically significant improvements compared to the other solution approaches featured in the computational experiments.

Automated summary

The paper proposes a Reinforcement Learning Iterated Local Search heuristic method for scheduling multiple additive manufacturing machines, which outperforms other solution approaches significantly in computational experiments.