Labeling in CLP(FD) with Evolutionary Programming

Constraint logic programming over finite integer domains allows a declarative and flexible statement of combinatorial optimization problems. The paradigm used is "constraint and generate". Constraints prune in advance the search space, and then, a enumeration phase, also called labeling, and a search strategy are needed to find the optimal solution. In this paper we introduce the integration of evolutionary algorithms, a well known computing paradigm, and the CLP(FD) paradigm. We have designed a system that enhances CLP with techniques based in evolutionary programming, allowing to solve a constrained optimization problems without the need of programming an explicit and exhaustive enumeration and specifying a strategy to find the optimal solution. The paper describes the algorithms used to implement the evolutionary program, and also the design details of the genetic operators. Finally, we present an example of the operation of the prototype of the system, which has been implemented in ECLiPSe.