Scheduling Lacquer Production by Reachability Analysis – A Case Study

Abstract

In this paper we describe a case study on lacquer production scheduling that was performed in the European IST-project AMETIST and was provided by one of the industrial partners. The approach is to derive schedules by means of reachability analysis: with this technique the search mechanism of model checkers, in our case here Uppaal, is used to find feasible or optimal schedules. The advantage of this approach is that the expressiveness of timed automata allows to model scheduling problems of different kinds, unlike many classical approaches, and the problem class is robust against changes in the parameter setting. To fight the typical state space explosion problem a number of standard heuristics have to be used. We discuss the difficulties when modelling an industrial case of this kind, describe the experiments we performed, the heuristics used, and the techniques applied to allow to optimize costs (storage costs, delay costs, etc.) while searching for schedules.

Synthesis and stochastic assessment of cost-optimal schedules

“…an optimal schedule. However, it was shown that this technique, extended by suitable heuristics, allows to find good schedules in very short time [5,6].…”

“…In earlier work [5,6] we experimented with several heuristics. In the current context, we used only greediness as strategy and restrict the number of orders that may be processed at any point in time.…”

“…The tool we use is the model-checker UPPAAL CORA, a version of UPPAAL which allows to attach costs to states of the considered model and which allows to find schedules that minimize the costs. The principles of this approach have been described in several publications [8,5,6].…”

Complexity in Simplicity: Flexible Agent-Based State Space Exploration

“…Costs are incurred during machine use, cleaning and storage (if recipes are fulfilled before the deadline). The Axxom case has been studied for schedulability with timed automata in [5] and for optimization in [13]. For a thorough introduction to the Axxom case study, we refer the reader to either of the two papers.…”

“…PTA have proven useful for a wide range of different search problems such as model checking, [7], and scheduling, [1,5,18]. The diversity of applications and generality of the modelling language of PTA suggests that no single search algorithm is superior and, thus, could benefit from an agent-based approach to search.…”