Local search optimization tools require writing a search procedure with the implementation of various heuristics. To make this stage more declarative and easily maintainable, a powerful combinator language was developed at CETIC and implemented within the CBLS engine of the Oscar open source library. This contribution will be presented by Renaud De Landtsheer at the 11th Metaheuristics International Conference (MIC 2015) to be held in Agadir from 7 to 10 June.
This work presents a declarative framework to define local search procedures. It is achieved by combining neighborhoods through combinators which describes when neighborhoods need to be explored. They also allow to manage other aspects of search procedures such as stopping criteria, management of the solution, and meta-heuristics (eg. Simulated annealing). Our approach allows the developer to focus on these high-level concepts without worrying about their efficient implementation. Therefore, the development, understanding, communication, and testing procedures are more effective, and ultimately produce better search procedures.