Reconfigurable Scheduling through Discrete-Event Systems

Manufacturing systems and other highly commercially valuable systems of a similar structure remain only partially optimised; there have been few successful attempts at real-time, global optimisation of complex systems as a result of the inherent combinatorial state explosion. The focus of this research is to investigate and develop a theoretical framework for reconfigurable scheduling and control of such systems through the use of Discrete-Event Systems (DES) within the broader context of “Industrie 4.0” with a focus on manufacturing applications. The work presents a wide ranging overview of the existing approaches towards scheduling and discrete control of distributed, resource-allocation systems, the implementation of such systems within the contemporary Information Technology (IT) landscape and some theoretical fields that neighbour the work. The structure of a DES for scheduling problems is defined as a special case of a generative Markovian transition system. A full-scale industrial case study from the aerospace industry is modelled. The system is formalised into a parallel computer program with a Monte-Carlo sampling approach to illustrate the speed and effectiveness of the technique in sampling-based makespan minimisation in complex scheduling problems. Although approach is anytime-optimal, ideal for implementation into distributed computation, it does not intensify the search into high performing regions. The principles and appropriateness of existing metaheuristics are discussed, a simple explore-exploit algorithm called Discrete-Event Trajectory Mutation specifically designed for search intensification for sampling-based Discrete-Event Processes (DEP) is shown. A new scheduling problem driven by industrial requirements called “satisfaction-over-time” is defined using original theory, followed by an approach for formally representing these problems and a technique for solving them through computer optimisation and search. Finally, a technique for automatic construction of DES models for the search and optimisation of manufacturing system designs is presented. Extensive plans for further work are given along with profitable areas of further study.