Development and implementation of a framework for Model Predictive Control of discrete-events manufacturing systems

Model Predictive Control (MPC) is widely recognized for its ability to optimize complex systems, but its application to discrete-event manufacturing systems often demands detailed equation-based modelling, limiting flexibility and adaptability. This thesis introduces an open-source software framework that automates the synthesis of MPC controllers based solely on plant topology, enabling a natural, topology-driven approach that bypasses traditional mathematical formulations. This framework makes MPC more accessible for dynamic manufacturing environments by automating controller design without extensive manual setup. Built on free software principles, the framework is developed in Python and employs Google’s CP-SAT solver to handle the discrete optimization challenges. By leveraging CP-SAT’s constraint programming capabilities, the framework efficiently manages the discrete and nonlinear dynamics, allowing for real-time adaptability and control. Its object-oriented design provides modularity, making it easy for users to define and extend system components, while Python’s versatility ensures compatibility with other opensource tools. Simulation results demonstrate that the synthesized MPC controllers effectively optimize production speed, resource utilization, and constraint adherence across various manufacturing scenarios, requiring minimal user intervention. The framework’s performance shows it’s well-suited for industries that need to quickly reconfigure manufacturing lines to keep up with shifting production demands. This thesis shows that a topology-driven, open-source MPC framework can significantly reduce the complexity of controller design for discrete-event systems. This framework contributes to advancing automation in manufacturing, providing a scalable and free solution to enhance control performance while maintaining adaptability in complex, discrete-event environments.