A reactive approach for solving constraint satisfaction problems

We propose in this paper a multi-agent model for solving a class of Constraint Satisfaction Problems: the assignment problem. Our work is based on a real-world problem, the assignment of land-use categories in a farming territory, in the north-east of France. This problem exhibits a function to optimize, while respecting a set of constraints, both local (compatibility of grounds and land-use categories) and global (ratio of production between land-use categories). We developed a model using a purely reactive multi-agent system that builds its solution upon conflicts that arise during the resolution process. In this paper, we present the reactive modelling of the problem solving and experimental results from two points of view: the efficiency of the problem being solved and the properties of the problem solving process.