Abstract
In this paper we introduce a complex scheduling problem that arises in a real-world industrial test laboratory, where a large number of activities has to be performed using qualified personnel and specialized equipment, subject to time windows and several other constraints. The problem is an extension of the well-known Resource-Constrained Project Scheduling Problem and features multiple heterogeneous resources with very general availability restrictions, as well as a grouping phase, where the jobs have to be assembled from smaller units. We describe an instance generator for this problem and publicly available instance sets, both randomly generated and real-world data. Finally, we present and evaluate different metaheuristic approaches to solve the scheduling subproblem, where the assembled jobs are already provided. Our results show that Simulated Annealing can be used to achieve very good results, in particular for large instances, where it is able to consistently find better solutions than a state-of-the-art constraint programming solver within reasonable time.
Highlights
Project scheduling problems appear in countless variations wherever multiple activities have to be scheduled and assigned resources of some kind, subject to various constraints
We have introduced the new problem Test Laboratory Scheduling Problem (TLSP), and its subproblem TLSP-S, which are complex extensions to existing Resource-Constrained Project Scheduling Problem (RCPSP) variants based on real-world requirements
A complexity analysis via reduction from RCPSP shows that even finding a feasible solution for either of these two problems is already NP-hard
Summary
Project scheduling problems appear in countless variations wherever multiple activities have to be scheduled and assigned resources of some kind, subject to various constraints. Besides several well-known features of (extensions of) the RCPSP from the literature, such as multiple execution modes and time windows, the TLSP features additional constraints imposed by the real-world problem setting. In real-life practice, it is commonly the case that the grouping of tasks into jobs is already known and only a solution for the scheduling part of the problem is required. This gives rise to a restricted problem variant we denote as TLSP-S, which has a (fixed) list of jobs as additional input, but otherwise follows the same restrictions as TLSP.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have