Modelling Local Search in a Knowledge Base System

In this paper we present how the basic building blocks of local search approaches—problem constraints, neighbourhood moves, objective function, move evaluations—can be modelled declaratively using FO (\(\cdot \)), an extension of first order logic. We extend the Knowledge Base System IDP with three built-in local search heuristics, namely first improvement, best improvement and tabu search, which take those building block specifications as input and execute local search accordingly. To demonstrate the framework, three neighbourhood moves for three different problems are modelled and tested.