Journal of Computational Physics 159, 139–171 (2000) doi:10.1006/jcph.1999.6413, available online at http://www.idealibrary.com on Record Breaking Optimization Results Using the Ruin and Recreate Principle Gerhard Schrimpf, ∗ Johannes Schneider,† Hermann Stamm-Wilbrandt, ∗ and Gunter Dueck ∗ ∗ IBM Scientific Center Heidelberg, Vangerowstr. 18, D-69115 Heidelberg, Germany; †Faculty of Physics, University of Regensburg, D-93040 Regensburg, Germany E-mail: schrimpf@de.ibm.com, Johannes.Schneider@physik.uni-regensburg.de, stammw@de.ibm.com, dueck@de.ibm.com Received December 24, 1998; revised November 16, 1999 A new optimization principle is presented. Solutions of problems are partly, but sig- nificantly, ruined and rebuilt or recreated afterwards. Performing this type of change frequently, one can obtain astounding results for classical optimization problems. The new method is particularly suited for more complex optimization problems (“dis- continuous” ones, problems with hard-to-find admissible solutions, problems with complex objectives or many constraints). The method is an all-purpose-heuristic. Numerical results are given for the Traveling Salesman Problem, for the Vehicle Routing Problem with time windows, and for network optimization. Numerical evi- dence for the quality of the proposed principle is given. For most of the instances of a research library of problems, the ruin and recreate (R&R) implementation achieved the best published results. For many instances, better or much better solutions could be found. c  2000 Academic Press Key Words: combinatorial optimization; Monte Carlo, threshold accepting; global optimization; Traveling Salesman Problem; Vehicle Routing Problem; network op- timization. I. RUIN AND RECREATE, A FIRST LOOK AT THE PRINCIPLE Before we give a more systematic introduction, we want to give the reader a quick feeling for this new class of algorithms we introduce here. The basic element of our idea is to obtain new optimization solutions by a considerable obstruction of an existing solution and a following rebuilding procedure. Let’s look at a famous Traveling Salesman Problem which was often considered in the literature (PCB442 problem of Gr ¨ otschel [1–8]). Suppose we have found some roundtrip through all of the 442 cities like in Fig. 1. That’s our initial or current solution of the problem. We “ruin” now a significant part of the solution. That’s the easy part of it. When ruining the solution, think of a major 139 0021-9991/00 $35.00 Copyright c  2000 by Academic Press All rights of reproduction in any form reserved.