Using injection points in reformulation local search for solving continuous location problems

Abstract

Reformulation local search (RLS) has been recently proposed as a new approach for solving continuous location problems. The main idea, although not new, is to exploit the relation between the continuous model and its discrete counterpart. The RLS switches between the continuous model and a discrete relaxation in order to expand the search. In each iteration new points obtained in the continuous pha.se are added to the discrete formulation. Thus, the two formulations become equivalent in a limiting sense. In this paper we introduce the idea of adding 'injection points' in the discrete pha.se of RLS in order to escape a current local solution. Preliminary results are obtained on benchmark data sets for the multi-source Weber problem that support further investigation of the RLS framework.