Greedy approach and multi-criteria simulated annealing for the car sequencing problem

Abstract

In this paper we describe our approach to the car sequencing problem which was the topic of the ROADEF’05 challenge. A full description of the problem is given in a separate paper of this same issue of this journal. We decided to tackle the problem via simulated annealing. Our algorithm takes the three criteria in decreasing order of importance and optimize each time on a criterion without authorizing degradation of the objective on previously optimized criteria. There is a large panel of moves to choose from and we tried to choose the moves using dynamic probabilities which in some sense take into account the rate of success of each move for the given instance. We report the numerical results on the instances provided during that challenge.