Variable neighbourhood pump heuristic for 0-1 mixed integer programming feasibility

Abstract

In this paper we propose a new method for finding an initial feasible solution for Mixed integer programs. We call it "Variable neighborhood pump", since it combines ideas of Variable neighborhood branching and Feasibility pump heuristics. The proposed heuristic was tested on an established set of 83 benchmark problems proven to be difficult to solve to feasibility. The results are compared with those of IBM ILOG CPLEX 11.1, which already includes standard feasibility pump as a primal heuristic. With our approach we managed to obtain better initial objective function values than CPLEX on 63 test instances, within similar average computational time.