Detailed planning in a milk powder factory: a CSP modelling approach.

Constraint satisfaction techniques can be used to solve detailed planning problems in real situations (Dockx et al., 1997; Sadeh and Fox, 1996), a summary can be found in Wallace (1996). Application of constraint satisfaction techniques presumes a model and a solver. In other words: there has to be a model of the planning situation, which consists of variables, domains and constraints (a Constraint Satisfaction Problem (CSP)) and there has to be a solver (based on an algorithm or heuristic) which processes the model and produces tuples which satisfy the constraints. CSP research focuses mainly on improving algorithms and heuristics used to prune the search space (Sadeh and Fox, 1996; Kumar, 1992; Van Hentenryck et al., 1992). In this paper we will focus on the modelling aspect. We will identify and discuss important and obvious modelling choices, which are usually encountered when designing an object oriented constraint based model for a real world planning situation. For illustrating purposes we will construct several models of a milk powder factory. This factory has many characteristics that are relevant in food industry, which is the broader scope of our research. In the second section we will describe the factory, products that are made and processes that take place. In the third section we will describe the input of the detailed planning system and the desirable outputs. Various modelling choices and possible consequences are discussed in section four. The last section outlines the results of implementing the models using ILOG tools.