Constraint handling in evolutionary search: A case study of the frequency assignment

In this paper, we present the constraint handling techniques developed to tackle a real world optimization application: the frequency assignment problem (FAP) in cellular radio networks. The main goal of this application consists in finding assignments which minimize electromagnetic interference due to frequency reuse and minimize the number of frequencies used. An acceptable assignment of the FAP must satisfy a set of multiple constraints such as traffic constraints and interference constraints. We present how each type of constraint can play a different role in EAs. In particular, we show how co-station constraints can be used to reduce the search space efficiently. The effectiveness of the constraint handling techniques combined with a regeneration technique is tested on big and hard FAP instances. Results show that limiting the degree of unfeasible solutions is beneficial for this application.