Enhancing the effectiveness of joint production and maintenance scheduling based on a multi-agent system and a Pigouvian approach of externalities

Joint maintenance and production scheduling is very important and challenging in the smart manufacturing field. Multi-agent systems based on Contract-Net Protocol (CNP) have been proposed to solve this scheduling problem, to reduce complexity, increase flexibility, and enhance robustness. However, these multi-agent approaches use myopic behaviors which can cause loss of performance. This study aims to improve the CNP-based approach of joint production and maintenance scheduling by reducing loss of performance due to agent myopia. Based on a Pigouvian approach of externalities, a new multi-agent scheduling system RCEMP (Regulator, Customers, Environment, Maintainers, and Producers) was developed to improve the multi-agent system SCEMP (Supervisor, Customers, Environment, Maintainers, and Producers). We have implemented the SCEMP and RCEMP approaches with Python programming language, simulated them with a fixed set of problem instances and compared their performances with the Wilcoxon signed-rank tests. These comparisons show that the approach in terms of externalities improves significantly the performances of the CNP-based approach in terms of regular classical optimality criteria of scheduling, maintenance costs, and execution times.