Stable solving of CVRPs using hyperheuristics

In this paper we present a hill-climbing based hyperheuristic which is able to solve instances of the capacitated vehicle routing problem. The hyperheuristic manages a sequence of constructive-perturbative pairs of low-level heuristics which are applied sequentially in order to construct and improve partial solutions. We present some design considerations that we have taken into account to find the most promising sequence and allow the collaboration among low-level heuristics. Our approach has been tested using some standard state-of-the-art benchmarks and we have compared them with several well-known methods proposed in the literature. We have obtained, on average, stable and good quality solutions after solving various types of problems. Thus, we conclude that our collaborative framework is an interesting approach as it has proved to be: (1) able to adapt itself to different problem instances by choosing a suitable combination of low-level heuristics and (2) capable of preserving stability when solving different types of problems.